{"$about_the_source": {"collected_by_the_market": "Free programs made by OTHER people. Nobody gave them to us, we do not sell them and we earn nothing from them. They are here because they are useful. We have also run 6 of them ourselves: those carry `evidence.label` = MEASURED, with the machine and the date on the figure. Another 63 carry TESTED - we used them and we say what we found, but we publish no number and we say why. The remaining 347 have nothing from us, and that is exactly what they say.", "listed_by_a_seller": "A product on the shelf, under a seller's handle, with whatever measurement exists for it beside it.", "seller": "Listed by a seller registered here. The shelf is not open to outside sellers yet, so this is zero for now, and we would rather say so than decorate it."}, "count": 431, "collected_count": 416, "from_registered_sellers": 0, "listed_count": 15, "products": [{"description": "An AI that runs entirely on your Android phone. Photograph a page or open a PDF and ask about it; nothing leaves the phone, and it works in airplane mode. Android 13+, 8 GB RAM. One payment, no subscription.", "kind": "app", "id": "zember-ember", "name": "Zember Ember — Qwen3.5 4B, phone", "page": "https://marketaiverse.com/zember-ember", "price": {"currency": "EUR", "value": 39.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Ministral 3 3B Q4_K_M (2.0 GB), for graphics cards with 4 GB or more (GTX 970, GTX 1050 Ti). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 184.2 tok/s. One key, for this model only.", "kind": "software", "id": "zember-spark", "name": "Zember Spark — Ministral 3 3B", "page": "https://marketaiverse.com/zember-spark", "price": {"currency": "EUR", "value": 39.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Qwen3.5 4B Q6_K (3.3 GB), for graphics cards with 6 GB or more (GTX 980 Ti, GTX 1060 6GB, RTX 2060). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 129.3 tok/s. One key, for this model only.", "kind": "software", "id": "zember-glow", "name": "Zember Glow — Qwen3.5 4B", "page": "https://marketaiverse.com/zember-glow", "price": {"currency": "EUR", "value": 49.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Qwen3.5 9B Q4_K_M (5.3 GB), for graphics cards with 8 GB or more (GTX 1070, GTX 1080, RTX 2070). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 97.4 tok/s. One key, for this model only.", "kind": "software", "id": "zember-flame", "name": "Zember Flame — Qwen3.5 9B", "page": "https://marketaiverse.com/zember-flame", "price": {"currency": "EUR", "value": 59.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Ministral 3 14B Q4_K_M (7.7 GB), for graphics cards with 10 GB or more (RTX 3080 10GB, GTX 1080 Ti, RTX 2080 Ti). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 64.2 tok/s. One key, for this model only.", "kind": "software", "id": "zember-torch", "name": "Zember Torch — Ministral 3 14B", "page": "https://marketaiverse.com/zember-torch", "price": {"currency": "EUR", "value": 69.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Phi-4 14B Q5_K_M (9.7 GB), for graphics cards with 12 GB or more (RTX 3060 12GB, RTX 4070, RTX 5070 12GB). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 56.8 tok/s. One key, for this model only.", "kind": "software", "id": "zember-blaze", "name": "Zember Blaze — Phi-4 14B", "page": "https://marketaiverse.com/zember-blaze", "price": {"currency": "EUR", "value": 79.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Qwen3.8 27B UD-IQ4_XS (13.3 GB), for graphics cards with 16 GB or more (RTX 4080 16GB, RTX 5080 16GB). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 7.0 tok/s. One key, for this model only.", "kind": "software", "id": "zember-forge", "name": "Zember Forge — Qwen3.8 27B", "page": "https://marketaiverse.com/zember-forge", "price": {"currency": "EUR", "value": 89.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Granite 4.1 30B Q5_K_M (19.1 GB), for graphics cards with 24 GB or more (RTX 3090, RTX 4090, Tesla P40 24GB). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 5.1 tok/s. One key, for this model only.", "kind": "software", "id": "zember-furnace", "name": "Zember Furnace — Granite 4.1 30B", "page": "https://marketaiverse.com/zember-furnace", "price": {"currency": "EUR", "value": 99.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Qwen3.6 35B-A3B Q6_K (28.8 GB), for graphics cards with 32 GB or more (RTX 5090 32GB). Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 24.9 tok/s. One key, for this model only.", "kind": "software", "id": "zember-crucible", "name": "Zember Crucible — Qwen3.6 35B-A3B", "page": "https://marketaiverse.com/zember-crucible", "price": {"currency": "EUR", "value": 119.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Qwen3.6 35B-A3B Q4_K_M (20.8 GB), for a card with 6 GB or more plus enough system RAM: most of the brain sits in RAM, by design. Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 29.8 tok/s. One key, for this model only.", "kind": "software", "id": "zember-inferno", "name": "Zember Inferno — Qwen3.6 35B-A3B", "page": "https://marketaiverse.com/zember-inferno", "price": {"currency": "EUR", "value": 99.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"description": "Qwen3.5 122B-A10B IQ3_S (43.4 GB), for a card with 12 GB or more plus enough system RAM: most of the brain sits in RAM, by design. Reads every page of a contract and cites the page for every point; nothing leaves your computer. Contract test: 12 of 12 right; we measured 13.6 tok/s. One key, for this model only.", "kind": "software", "id": "zember-wildfire", "name": "Zember Wildfire — Qwen3.5 122B-A10B", "page": "https://marketaiverse.com/zember-wildfire", "price": {"currency": "EUR", "value": 149.0}, "evidence": {"$what_it_means": "TESTED = we started the thing and put it to work. REAL = a number was taken on the exact configuration promised.", "label": "TESTED", "note": "The models marked on the product page were run here, with the machine and the date next to each number. The speed shown for each package, though, is for the BUYER'S card, and it is computed — from an efficiency figure that was also measured here, not from a textbook constant."}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "zember"}, {"capabilities": {"needs": {"from_the_customer": "the manufacturer's original manual, as a PDF", "from_us": "runs on our machine; the customer installs nothing"}, "outputs": ["PDF", "Markdown"], "inputs": ["PDF"], "languages": ["RO"], "called_via": ["human"], "we_know_this_too": {"diacritics": "10 of 10 Romanian diacritics kept, and the cedilla look-alikes straightened", "measured_on": "Contabo, 12 vCPU, 48 GB, no graphics card, 2026-09-13", "pages_checked": 60, "text_retention": "101.5% on the 60-page document (it goes over 100 because the tool adds page markers)", "where_it_breaks": "a page holding ONLY a table gets dropped: it has no Romanian words to recognise, so its language score comes out 0.0000 against a threshold of 0.012"}}, "description": "A manual in Chinese or English comes back as a PDF in your language, with the diagrams in place and the tables intact. Not a text translation — a rebuilt document.", "kind": "tool", "id": "manuale", "name": "Technical manuals, translated and typeset", "page": "https://marketaiverse.com/product-manuals", "price": {"note": "priced by manual size", "value": null}, "evidence": {"label": "MEASURED", "note": "Measured here. The number and the method are on the product page."}, "standard_benchmark": {"what_failed": [{"what_happened": "pages holding nothing but a table were dropped: 0 of 3 kept, with no error", "entry": "2-tabele.pdf"}], "date": "2026-09-13", "sheet": "https://marketaiverse.com/sheet-manuals", "machine": "Contabo (12 vCPU, 48 GB, no graphics card)", "note": "Two of the seven inputs MUST fail (6 and 7). Both did, so the check can still say no.", "failed": 1, "set": "documents-v1", "passed": 6}, "state": "for_sale", "source": "listed_by_a_seller", "seller": "xedra"}, {"description": "Extract the tables and fields from a PDF, a scanned page or a photo: tables as CSV, fields such as number, date and totals. Also the MCP tool `extract_tables`. The file is deleted once read.", "kind": "tool", "id": "extragere", "name": "Invoice and PDF to CSV", "page": "https://marketaiverse.com/extract", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "No measurement published for this one."}, "state": "free", "source": "listed_by_a_seller", "seller": "xedra"}, {"description": "For large catalogues: descriptions written from the spec sheet, not invented. Works across thousands of products without getting bored.", "kind": "tool", "id": "descrieri", "name": "Product descriptions and search copy", "price": {"note": "per item, coming", "value": null}, "evidence": {"label": "NOT_MEASURED", "note": "The standard benchmark has not been run on this yet, so no number is published."}, "state": "coming", "source": "listed_by_a_seller", "seller": "xedra"}, {"description": "Images generated to order. For shops, for posts, for presentations.", "kind": "images", "id": "imagini", "name": "Illustrations, logos, posters", "price": {"note": "per image, coming", "value": null}, "evidence": {"label": "NOT_MEASURED", "note": "The standard benchmark has not been run on this yet, so no number is published."}, "state": "coming", "source": "listed_by_a_seller", "seller": "xedra"}, {"address": "https://github.com/ggml-org/llama.cpp", "what_it_is": "inference engine", "description": "The engine under almost every tool that runs models at home. Runs GGUF files on CPU, CUDA, Metal, ROCm and Vulkan. Everything happens on your own machine: it works with no internet at all, and nothing is sent to the cloud. Made by ggml.ai and contributors.", "kind": "run locally", "id": "u-llamacpp", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ggml-org/llama.cpp/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "llama.cpp", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$about_the_machine": "This number is about the machine named in `machine`, not about the tool in general. A different card gives a different number, and we do not have it.", "details": {"VRAM": "2709 MiB of 12227", "context": "8192", "model": "Qwen3.5-0.8B, Q8_K_XL (1.19 GB)", "startup": "a few seconds"}, "what_we_measured": "Local chat throughput, small model, all layers on the GPU", "figure": "163.9 tokens/second", "how": "llama-server with --n-gpu-layers 99 --ctx-size 8192 --batch-size 512. One request through the HTTP API, timed end to end; the token count comes from the engine's own usage report, not estimated from characters.", "date": "2026-09-13", "worth_knowing": "Newer models think before answering, and the whole answer goes into `reasoning_content` while `content` comes back EMPTY — with no error anywhere. Pass --reasoning off unless you want that, or your users will see a blank box.", "label": "MEASURED", "the_whole_chain": ["llama.cpp"], "machine": "RTX 5070 12GB, Windows 11, CUDA"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://ollama.com", "what_it_is": "local model runner", "description": "One command and the model is on your machine, with an OpenAI-compatible endpoint. You pull models by name, the way you pull images with Docker. After that it runs offline, and your prompts never leave the computer. Made by Ollama.", "kind": "run locally", "id": "u-ollama", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Ollama", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-06 to 2026-08", "why_no_figure": "The numbers we have are for the models we ran under it, not for it.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "Ran it for months as our model runner. Excellent to start with; we moved off it for the big quantised model, where we needed control over expert offloading."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/vllm-project/vllm", "what_it_is": "serving engine", "description": "A serving engine built for throughput: it batches many requests onto one model instead of taking them one at a time. Made by proiectul vLLM.", "kind": "run locally", "id": "u-vllm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/vllm-project/vllm/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "vLLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://www.nomic.ai/gpt4all", "what_it_is": "desktop app", "description": "Runs models on ordinary laptops, including machines with no graphics card, and can index a folder on your computer — private AI for your own documents, offline, with nothing uploaded to a cloud service. Made by Nomic AI.", "kind": "run locally", "id": "u-gpt4all", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "GPT4All", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Its speed is the speed of whatever model you load into it, so a number here would belong to the model, not to the app.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. It is a desktop app rather than a server, so it is the one thing on this shelf you can start without opening a terminal - which matters more than it sounds if you are handing a local model to someone who does not want to learn one."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://lmstudio.ai", "what_it_is": "desktop app", "description": "A desktop app for finding, comparing and talking to local models. Good when you want to see what exists rather than build something. The models run on your own computer, offline, with and it does not send your data to the cloud. Made by LM Studio.", "kind": "run locally", "id": "u-lmstudio", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "LM Studio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Its speed is the speed of the model you load into it.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. A desktop app rather than a server, so it is one of the few things here you can start without a terminal."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://n8n.io", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "JSON"]}, "what_it_is": "workflow automation", "description": "A workflow tool you keep on your own server, wiring models to hundreds of other programs. Made by n8n.", "kind": "agents", "id": "u-n8n", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/n8n-io/n8n/HEAD/LICENSE.md", "family": "custom", "id": "custom"}, "name": "n8n", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://flowiseai.com", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "visual agent builder", "description": "A canvas you drag on to build agents and model workflows without writing code. Made by FlowiseAI.", "kind": "agents", "id": "u-flowise", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/FlowiseAI/Flowise/HEAD/LICENSE.md", "family": "permissive", "id": "Apache-2.0"}, "name": "Flowise", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://dify.ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "PDF"]}, "what_it_is": "LLM app platform", "description": "A place to build apps on top of models: prompts, data and watching how they behave, in one place. Made by Dify.", "kind": "agents", "id": "u-dify", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/langgenius/dify/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Dify", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://www.langchain.com", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "code"]}, "what_it_is": "agent library", "description": "The library agents get written with, in code. The most widespread one, and where most of the examples you find start. Made by LangChain.", "kind": "agents", "id": "u-langchain", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/langchain-ai/langchain/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LangChain", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://www.langchain.com/langgraph", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "code"]}, "what_it_is": "agent framework", "description": "The agent drawn as a graph: states, steps and places where a human steps in. For when the flow no longer fits in a straight line. Made by LangChain.", "kind": "agents", "id": "u-langgraph", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/langchain-ai/langgraph/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LangGraph", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://www.crewai.com", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "multi-agent framework", "description": "Several agents that split the work between them, each with its own role. Made by CrewAI.", "kind": "agents", "id": "u-crewai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/crewAIInc/crewAI/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "CrewAI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://haystack.deepset.ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "PDF"]}, "what_it_is": "search framework", "description": "A framework for search and answering from your own documents. Connects to most of the vector databases below. Made by deepset.", "kind": "agents", "id": "u-haystack", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/deepset-ai/haystack/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Haystack", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co", "what_it_is": "model & dataset hub", "description": "Where the open-weight models and the datasets actually live. You take them from here before running them with any of the tools here. Made by Hugging Face.", "kind": "models", "id": "u-hf", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Hugging Face", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "used continuously through 2026", "why_no_figure": "It is a hub and a client library — timing it would measure our connection, not the tool.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11, and our Linux server", "verdict": "We pull every model through it, every week. It does its one job without drama."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://www.comfy.org", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image", "video"], "inputs": ["text", "image"]}, "what_it_is": "node editor", "description": "A node editor for diffusion models. Images, video and audio all come out of the same place, by wiring boxes together. Made by Comfy Org and contributors.", "kind": "media", "id": "u-comfyui", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/Comfy-Org/ComfyUI/HEAD/LICENSE", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "ComfyUI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "used continuously, last measured run 2026-09-06", "why_no_figure": "It is a runner, not a model. Timing it measures whatever you put inside it, so the number would belong to the model, not to this.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "The engine behind every video number on this shelf. One thing worth knowing before you trust it: a preset in its graph can silently replace your prompt instead of enriching it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://stability.ai", "what_it_is": "image model", "description": "The family of image models almost everything people run at home started from. Usually used through ComfyUI. Made by Stability AI.", "kind": "media", "id": "u-sd", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Stable Diffusion", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed. Image speed depends so much on steps, sampler and resolution that one number would mislead more than it informs.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Our own image work ended up on FLUX and on the small distilled models instead, so we have no quality claim to make here beyond having used it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Wan-Video", "what_it_is": "video model", "description": "An open-weight video model. The small build runs on eight-gigabyte cards, with offloading to system memory. Made by Alibaba.", "kind": "media", "id": "u-wan", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Wan 2.2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-07-11", "why_no_figure": "The generation time we wrote down was for one specific build and length, and this card covers the family — we will not stretch one number across versions.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. On natural footage the 480p output plus an upscale held up; on fantasy prompts we rejected it. The 14B NF4 build peaked at 11.6 GB on a 12 GB card, so it fits, but with nothing to spare."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Lightricks/LTX-Video", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["text", "image"]}, "what_it_is": "video model", "description": "A video model that produces picture and sound together, in the same pass. Made by Lightricks.", "kind": "media", "id": "u-ltx", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Lightricks/LTX-Video/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "LTX-Video", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$about_the_machine": "This number is about the machine named in `machine`, not about the tool in general. A different card gives a different number, and we do not have it.", "details": {"1080p_instead_of_720": "+3.4 min", "480p_instead_of_720": "-2.8 min", "asezarea_masurata": "5 s, 720p, `medium`, model already warm", "why": "241 frames instead of 121, and attention grows with the square of the length. The transformer is 14.6 GB and the card has 12, so the machine drags the model through system RAM. This is not slow, it is not feasible.", "light_instead_of_medium": "-0.7 min", "ultra_instead_of_medium": "+1.0 min", "ten_seconds": "never finished on this card. Measured three times: stopped at 39 min, retried past 80 min with ComfyUI holding 42 GB of system RAM, and 10 s at 1080p `ultra` went past 150 min."}, "what_we_measured": "LTX-2.5 text-to-video, end to end through the server — not the model call alone, but the whole wait a person actually sits through", "figure": "9.8 min for a 5-second clip", "how": "Timed through our own server, cold model excluded, one setting changed at a time against a fixed reference run (8 runs instead of a full 18-run matrix — the formula is additive, so one lever at a time plus a check run is enough).", "date": "2026-09-06", "worth_knowing": "In ComfyUI's LTX-2 graph, the `ultra` preset used to route the prompt through a Gemma rewriting node (`TextGenerateLTX2Prompt`). That node does not enrich your prompt, it REPLACES it: 'a red fox walking through tall grass at golden hour' came back as a crowd of people in golden crowns. The grass and the light survived; the subject did not. A benchmark that compares minutes will never catch this — you have to watch the clip.", "label": "MEASURED", "the_whole_chain": ["ComfyUI", "LTX-Video"], "machine": "RTX 5070 12GB, Windows 11, CUDA"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Tencent-Hunyuan", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["text", "image"]}, "what_it_is": "video model", "description": "Another open-weight video model, one of the large ones. Made by Tencent.", "kind": "media", "id": "u-hunyuanvideo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/Tencent-Hunyuan/HunyuanVideo/HEAD/LICENSE.txt", "family": "custom", "id": "custom"}, "name": "HunyuanVideo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-07-11", "why_no_figure": "It never completed a clean run on this card, so any time we wrote down would be the time of a failure, not of the tool.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "Peaked at 12.3 GB on a 12 GB card and never finished a clean run here. That is not a verdict on quality: it simply did not fit."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/zai-org/CogVideo", "what_it_is": "video model", "description": "A video model that runs on more modest hardware than the big ones. Made by Zhipu AI / THUDM.", "kind": "media", "id": "u-cogvideo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/THUDM/CogVideo/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "CogVideoX", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-08", "why_no_figure": "We did not time it, and we would rather say nothing than publish a figure from memory.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. It did not become part of our video pipeline — LTX-Video is what we settled on — so we have no quality claim to make beyond that."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/genmoai/mochi", "what_it_is": "video model", "description": "An open-weight video model from the recent wave. Made by Genmo.", "kind": "media", "id": "u-mochi", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/genmoai/models/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Mochi 1", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-08", "why_no_figure": "Not timed. The verdict is a choice we made, not a measurement.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it as an image-to-video candidate. It did not go further here; LTX-Video is the engine our own work runs on."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openai/whisper", "what_it_is": "speech to text", "description": "Turns sound into text: transcription, translation and timestamps. Runs entirely on your own machine, with no audio leaving it. Made by OpenAI.", "kind": "media", "id": "u-whisper", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openai/whisper/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Whisper", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-09-14", "why_no_figure": "The timing we have belongs to the runtime we used, not to this implementation, and we do not move a number from one card to another because the model name matches.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11, CUDA", "verdict": "We ran the large-v3 weights on Romanian speech and they came back correct: diacritics right, company names kept, no invented sentences. We ran them through faster-whisper rather than the reference implementation, so the speed on the neighbouring card is the one we can point at."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://www.trychroma.com", "what_it_is": "vector database", "description": "The easiest vector database to start: one install and it runs. Good for starting out and for trying things. Made by Chroma.", "kind": "memory", "id": "u-chroma", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Chroma", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://qdrant.tech", "what_it_is": "vector database", "description": "A vector search engine written in Rust, with fast filtering and a small memory footprint. Made by Qdrant.", "kind": "memory", "id": "u-qdrant", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Qdrant", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://weaviate.io", "what_it_is": "vector database", "description": "A database built for AI, with embedding built in and hybrid search. Made by Weaviate.", "kind": "memory", "id": "u-weaviate", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Weaviate", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://milvus.io", "what_it_is": "vector database", "description": "A vector database that spreads across machines and holds billions of vectors. Made by Zilliz.", "kind": "memory", "id": "u-milvus", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Milvus", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://lancedb.com", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["vector", "text"]}, "what_it_is": "vector database", "description": "A vector database that lives in files, next to your data, with no separate server. Made by LanceDB.", "kind": "memory", "id": "u-lancedb", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/lancedb/lancedb/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "LanceDB", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pgvector/pgvector", "what_it_is": "vector database", "description": "An extension for PostgreSQL that keeps vectors next to your ordinary data. If you already run Postgres, this is the shortest path. Made by pgvector and contributors.", "kind": "memory", "id": "u-pgvector", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/pgvector/pgvector/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "pgvector", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://vespa.ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "vector"]}, "what_it_is": "search and vectors", "description": "A search and serving engine that does vectors and classic search together, at scale. Made by Vespa.ai.", "kind": "memory", "id": "u-vespa", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/vespa-engine/vespa/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Vespa", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/docling-project/docling", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON"], "inputs": ["PDF", "DOCX", "PPTX", "image"]}, "what_it_is": "document parser", "description": "Reads PDF, DOCX, PPTX and images and gives back clean structured Markdown or JSON, keeping tables and reading order. Made by IBM Research. Runs locally, so private and legal documents are parsed offline, without being sent to a cloud service.", "kind": "documents", "id": "u-docling", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/docling-project/docling/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Docling", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/datalab-to/marker", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON", "HTML"], "inputs": ["PDF"]}, "what_it_is": "PDF to markdown", "description": "Turns PDF into Markdown, JSON or HTML, with tables, equations and page layout preserved.", "kind": "documents", "id": "u-marker", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/datalab-to/marker/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Marker", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "It is not in our published document comparison, and we do not publish a figure we cannot point at.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it on our own documents, converting PDFs to markdown."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/opendatalab/MinerU", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON"], "inputs": ["PDF"]}, "what_it_is": "document extraction", "description": "Extracts text, tables, formulas and images from PDFs, aimed at scientific and technical documents.", "kind": "documents", "id": "u-mineru", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/opendatalab/MinerU/HEAD/LICENSE.md", "family": "custom", "id": "custom"}, "name": "MinerU", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Same reason as Marker: not in the published comparison, so no number we can stand behind.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it on our own documents. It is the one we reach for when a PDF has layout worth keeping rather than just text worth pulling."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Unstructured-IO/unstructured", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "text"], "inputs": ["PDF", "DOCX", "HTML", "email"]}, "what_it_is": "document preprocessing", "description": "Turns many document formats into clean chunks ready for a model or a vector database.", "kind": "documents", "id": "u-unstructured", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Unstructured-IO/unstructured/HEAD/LICENSE.md", "family": "permissive", "id": "Apache-2.0"}, "name": "Unstructured", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/tesseract-ocr/tesseract", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "hOCR"], "inputs": ["image", "PDF"]}, "what_it_is": "OCR engine", "description": "The long-standing open OCR engine. Reads text out of images in over 100 languages. Runs offline on your own machine — private document scanning with nothing uploaded.", "kind": "documents", "id": "u-tesseract", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/tesseract-ocr/tesseract/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Tesseract OCR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-09-16", "why_no_figure": "We checked whether it got the values right, not how fast. On a scan, being right is the whole question.", "label": "TESTED", "machine": "Windows 11, CPU only", "verdict": "7 of 7 values came out correct on a scanned PDF with no text layer. It is what makes a scan readable at all."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/PaddlePaddle/PaddleOCR", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "OCR toolkit", "description": "OCR that handles many languages and rotated or curved text, with table and layout recognition on top.", "kind": "documents", "id": "u-paddleocr", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "PaddleOCR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/datalab-to/surya", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "OCR and layout", "description": "OCR with line-level detection, reading order and table recognition, across 90+ languages.", "kind": "documents", "id": "u-surya", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/datalab-to/surya/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Surya", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ocrmypdf/OCRmyPDF", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["PDF"], "inputs": ["PDF"]}, "what_it_is": "searchable PDF", "description": "Adds a text layer to a scanned PDF so it becomes searchable, without touching the original image.", "kind": "documents", "id": "u-ocrmypdf", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can link it into a closed product, but changes to THIS library must stay open.", "proof": "https://raw.githubusercontent.com/ocrmypdf/OCRmyPDF/HEAD/LICENSE", "family": "weak_copyleft", "id": "MPL-2.0"}, "name": "OCRmyPDF", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pymupdf/PyMuPDF", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image"], "inputs": ["PDF", "EPUB"]}, "what_it_is": "PDF library", "description": "Fast library for reading, writing and rendering PDF, XPS and EPUB.", "kind": "documents", "id": "u-pymupdf", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/pymupdf/PyMuPDF/HEAD/COPYING", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "PyMuPDF", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$about_the_machine": "This number is about the machine named in `machine`, not about the tool in general. A different card gives a different number, and we do not have it.", "details": {"diacritics": "10 of 10 Romanian characters kept", "the_large_file": "65,501 characters from 60+ pages in 0.0253 s", "the_broken_file": "refused it: 0 pages, 0 characters", "behaved_correctly": "7 of 7", "tables": "none returned as tables — it gives you the text, not the structure", "text_in_images": "lost, as expected — it does no OCR, and does not pretend to"}, "what_we_measured": "Text extraction from the seven public benchmark files, CPU only", "figure": "0.040 s for all seven files", "how": "Each file opened and fully read with fitz. Small files timed over 20 runs and averaged, the large one over 3 — a single run of a small file came back as 0.00 s, which is not a speed, it is the clock giving up. Same files, same machine, same minute as the pdfplumber measurement next to it.", "date": "2026-09-16", "worth_knowing": "It is roughly 40x faster than pdfplumber here, and that is the whole trade: you get text fast and you do NOT get tables as tables. If the document is half tables, the speed buys you nothing.", "label": "MEASURED", "the_whole_chain": ["PyMuPDF"], "machine": "Windows 11, 28 cores, CPU only"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/jsvine/pdfplumber", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "CSV", "JSON"], "inputs": ["PDF"]}, "what_it_is": "PDF table extraction", "description": "Pulls text, words, lines and tables out of a PDF with precise coordinates.", "kind": "documents", "id": "u-pdfplumber", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/jsvine/pdfplumber/HEAD/LICENSE.txt", "family": "permissive", "id": "MIT"}, "name": "pdfplumber", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$about_the_machine": "This number is about the machine named in `machine`, not about the tool in general. A different card gives a different number, and we do not have it.", "details": {"diacritics": "10 of 10 Romanian characters kept", "the_large_file": "65,501 characters from 60+ pages in 1.069 s", "the_broken_file": "REBUILT it completely — 10,574 characters, byte for byte the same as the intact original we compared against", "behaved_correctly": "7 of 7", "tables": "3 tables found and returned as tables", "text_in_images": "lost, as expected — it does no OCR"}, "what_we_measured": "Text extraction from the seven public benchmark files, CPU only", "figure": "1.57 s for all seven files", "how": "Each file opened, every page read, tables extracted per page. Timed the same way as PyMuPDF: 20 runs averaged for small files, 3 for the large one. The recovery claim was checked, not assumed: we compared the text against the undamaged version of the same document and it matched exactly.", "date": "2026-09-16", "worth_knowing": "Our own benchmark page called file 7 'no header, zero recoverable pages'. That was true of the tool we had, not of the file: pdfplumber recovers it whole. If you have a PDF that another library refuses, this is worth trying before you call the file dead — but check what comes out, because a library that returns invented text from a damaged file is worse than one that refuses.", "label": "MEASURED", "the_whole_chain": ["pdfplumber"], "machine": "Windows 11, 28 cores, CPU only"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/camelot-dev/camelot", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["CSV", "JSON"], "inputs": ["PDF"]}, "what_it_is": "PDF tables", "description": "Extracts tables from text-based PDFs into dataframes or CSV.", "kind": "documents", "id": "u-camelot", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/camelot-dev/camelot/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Camelot", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/markitdown", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown"], "inputs": ["PDF", "DOCX", "XLSX", "image", "audio"]}, "what_it_is": "anything to markdown", "description": "Converts Office files, PDFs, images, audio and HTML into Markdown for feeding a model. Made by Microsoft.", "kind": "documents", "id": "u-markitdown", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/markitdown/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MarkItDown", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Layout-Parser/layout-parser", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "document layout", "description": "Detects the layout of a document page: columns, figures, tables, captions.", "kind": "documents", "id": "u-layoutparser", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Layout-Parser/layout-parser/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "LayoutParser", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mindee/doctr", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "OCR library", "description": "End-to-end OCR built on PyTorch and TensorFlow: finds the text, then reads it.", "kind": "documents", "id": "u-doctr", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mindee/doctr/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "docTR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Ucas-HaoranWei/GOT-OCR2.0", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "Markdown"], "inputs": ["image"]}, "what_it_is": "general OCR model", "description": "An OCR model that also handles formulas, sheet music and charts, not only plain text.", "kind": "documents", "id": "u-gotocr", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "GOT-OCR2.0", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/facebook/nllb-200-distilled-600M", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "translation model", "description": "Translation across 200 languages, including many with little training data. Made by Meta.", "kind": "translation", "id": "u-nllb", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "NLLB-200", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/Helsinki-NLP", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "translation models", "description": "Hundreds of small language-pair translation models from the Helsinki-NLP group. One model per pair, so each is tiny.", "kind": "translation", "id": "u-opusmt", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "OPUS-MT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/argosopentech/argos-translate", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "offline translation", "description": "Offline translation with downloadable language packs, and a command-line tool. Offline translation: the text never leaves your machine, which matters for contracts and other private documents.", "kind": "translation", "id": "u-argos", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/argosopentech/argos-translate/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Argos Translate", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/google/madlad400-3b-mt", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "translation model", "description": "A translation model covering over 400 languages, trained on a documented public corpus.", "kind": "translation", "id": "u-madlad", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "MADLAD-400", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/LibreTranslate/LibreTranslate", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "translation API", "description": "A translation service you run yourself, with an HTTP API. Built on Argos. Self-hosted, so private and legal documents are translated without sending them to any cloud service.", "kind": "translation", "id": "u-libretranslate", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/LibreTranslate/LibreTranslate/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "LibreTranslate", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/facebook/seamless-m4t-v2-large", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio", "text"], "inputs": ["audio", "text"]}, "what_it_is": "speech and text translation", "description": "Translates between speech and text across many languages, in both directions. Made by Meta.", "kind": "translation", "id": "u-seamless", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "SeamlessM4T", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/SYSTRAN/faster-whisper", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "SRT", "VTT"], "inputs": ["audio"]}, "what_it_is": "speech to text", "description": "A reimplementation of Whisper that runs several times faster on the same hardware, using CTranslate2. Local and offline: private transcription of recordings that must not leave the building.", "kind": "sound", "id": "u-fasterwhisper", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/SYSTRAN/faster-whisper/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "faster-whisper", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$about_the_machine": "This number is about the machine named in `machine`, not about the tool in general. A different card gives a different number, and we do not have it.", "details": {"VRAM": "4073 MiB", "calitatea": "diacritics correct, company names kept, no invented sentences", "incarcarea_modelului": "6.8 s, once", "limba_detectata": "ro, confidence 1.00", "sunet": "39.4 s of Romanian speech (a factory owner talking, from a phone clip)", "transcris_in": "3.2 s"}, "what_we_measured": "Romanian speech-to-text, large-v3, float16, on the GPU", "figure": "12.5x faster than real time", "how": "faster-whisper large-v3, compute_type=float16, vad_filter=True, beam_size=5, condition_on_previous_text=False. Audio converted to 16 kHz mono WAV with FFmpeg first. Timed around list(transcribe(...)), because the generator does no work until it is consumed.", "date": "2026-09-14", "worth_knowing": "ctranslate2 (the engine under faster-whisper) needs cuBLAS and cuDNN and does NOT ship them. On Windows they are already on disk inside torch/lib — add that directory before importing, or the model loads fine and then dies on the first chunk of audio.", "label": "MEASURED", "the_whole_chain": ["yt-dlp", "FFmpeg", "faster-whisper"], "machine": "RTX 5070 12GB, Windows 11, CUDA"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ggml-org/whisper.cpp", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "SRT"], "inputs": ["audio"]}, "what_it_is": "speech to text", "description": "Whisper in plain C/C++, no Python needed. Runs on CPU, and on Apple silicon very well. Runs offline on your own machine, so recordings are never uploaded to a cloud service.", "kind": "sound", "id": "u-whispercpp", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ggml-org/whisper.cpp/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "whisper.cpp", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "We will not move a number from one runtime to another because the model name matches.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Same weights as the speech-to-text numbers on this shelf, a different runtime - ours were taken through faster-whisper."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/m-bain/whisperX", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON", "SRT"], "inputs": ["audio"]}, "what_it_is": "transcription with timing", "description": "Whisper plus word-level timestamps and speaker separation.", "kind": "sound", "id": "u-whisperx", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/m-bain/whisperX/HEAD/LICENSE", "family": "permissive", "id": "BSD-2-Clause"}, "name": "WhisperX", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed separately from the chain it sits in.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. What it adds over plain transcription is word-level timing, which is the part that matters if you are cutting video to speech."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OHF-Voice/piper1-gpl", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "text to speech", "description": "Fast local text-to-speech that runs even on a Raspberry Pi, with voices in many languages. Fully offline, on device, with nothing sent to a cloud service.", "kind": "sound", "id": "u-piper", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/OHF-Voice/piper1-gpl/HEAD/COPYING", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "Piper", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/hexgrad/Kokoro-82M", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "text to speech", "description": "A small text-to-speech model that sounds well above its size. Small enough to run locally and offline, with no cloud account.", "kind": "sound", "id": "u-kokoro", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Kokoro TTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/coqui/XTTS-v2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text", "audio"]}, "what_it_is": "voice cloning TTS", "description": "Text-to-speech that can copy a voice from a few seconds of audio, across languages.", "kind": "sound", "id": "u-xtts", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "XTTS-v2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed. What we were judging was whether the cloned voice held up, not how fast it arrived.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, for voice cloning."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/resemble-ai/chatterbox", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text", "audio"]}, "what_it_is": "text to speech", "description": "Open text-to-speech with voice cloning and control over how emphatic the delivery is.", "kind": "sound", "id": "u-chatterbox", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/resemble-ai/chatterbox/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Chatterbox", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/SWivid/F5-TTS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text", "audio"]}, "what_it_is": "text to speech", "description": "Text-to-speech that copies a voice from a short sample, with fast inference.", "kind": "sound", "id": "u-f5tts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/SWivid/F5-TTS/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "F5-TTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/facebookresearch/audiocraft", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text", "audio"]}, "what_it_is": "music generation", "description": "Generates music from a text description, optionally following a melody you give it. Made by Meta.", "kind": "sound", "id": "u-musicgen", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/facebookresearch/audiocraft/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MusicGen", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. ACE-Step is what we settled on for music here, so this one did not go further - that is a choice we made, not a defect we found."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/stabilityai/stable-audio-open-1.0", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "sound and music generation", "description": "Generates sound effects and short musical pieces from text.", "kind": "sound", "id": "u-stableaudio", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Stable Audio Open", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Same answer as MusicGen: we settled on ACE-Step, so this did not become part of our pipeline."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/adefossez/demucs", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["audio"]}, "what_it_is": "stem separation", "description": "Splits a song into vocals, drums, bass and the rest.", "kind": "sound", "id": "u-demucs", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/adefossez/demucs/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Demucs", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pyannote/pyannote-audio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["audio"]}, "what_it_is": "speaker diarization", "description": "Works out who is speaking and when, which turns a recording into a labelled conversation.", "kind": "sound", "id": "u-pyannote", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/pyannote/pyannote-audio/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "pyannote.audio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/snakers4/silero-vad", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["audio"]}, "what_it_is": "voice activity detection", "description": "Tells you which parts of an audio stream actually contain speech. Tiny and fast.", "kind": "sound", "id": "u-silerovad", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/snakers4/silero-vad/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Silero VAD", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["image", "text"]}, "what_it_is": "vision language model", "description": "Looks at an image or a document and answers questions about it, including reading text and understanding charts.", "kind": "vision", "id": "u-qwenvl", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Qwen2.5-VL", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed. The thing that decides this one is its licence, not its speed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Worth knowing before you build on it: in this family the non-commercial licence sits on the small size while the 7B and 72B are Apache-2.0 - so the one that fits a modest card is the one you cannot sell on. See the licence traps page."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenGVLab/InternVL", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["image", "text"]}, "what_it_is": "vision language model", "description": "A family of vision-language models, from small enough for a laptop up to large.", "kind": "vision", "id": "u-internvl", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenGVLab/InternVL/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "InternVL", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/m87-labs/moondream", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["image", "text"]}, "what_it_is": "small vision model", "description": "A tiny vision-language model built to run on modest hardware.", "kind": "vision", "id": "u-moondream", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/m87-labs/moondream/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "moondream", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We did NOT run this. We weighed it on paper — licence, architecture, how much it occupies, what encoder it drags along — and `why` is the reason that decision came out the way it did. It is not a review of how well it works, because we did not find that out.", "when": "2026-09-18", "why": "Two versions, two answers. `moondream2` is clean; `moondream3-preview` forbids paid products that compete with it. Check which one you pulled.", "label": "REVIEWED"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/microsoft/Florence-2-large", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["image", "text"]}, "what_it_is": "vision tasks", "description": "One small model for captioning, detection, segmentation and OCR. Made by Microsoft.", "kind": "vision", "id": "u-florence", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Florence-2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/facebookresearch/sam2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["mask", "JSON"], "inputs": ["image", "video"]}, "what_it_is": "segmentation", "description": "Cuts out any object in an image or video, from a click or a box. Made by Meta.", "kind": "vision", "id": "u-sam2", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/facebookresearch/sam2/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Segment Anything 2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ultralytics/ultralytics", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image", "video"]}, "what_it_is": "object detection", "description": "Detects, segments and tracks objects in images and video, in real time.", "kind": "vision", "id": "u-yolo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/ultralytics/ultralytics/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Ultralytics YOLO", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openai/CLIP", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["image", "text"]}, "what_it_is": "image-text matching", "description": "Puts images and text in the same space, so you can search pictures with words.", "kind": "vision", "id": "u-clip", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openai/CLIP/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "CLIP", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/IDEA-Research/GroundingDINO", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image", "text"]}, "what_it_is": "detection from words", "description": "Finds objects in an image described in plain words, without being trained on those classes.", "kind": "vision", "id": "u-groundingdino", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/IDEA-Research/GroundingDINO/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Grounding DINO", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/danielgatis/rembg", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "background removal", "description": "Removes the background from a photo in one call, leaving a clean cut-out with a transparent background. Works on people, products and animals.", "kind": "vision", "id": "u-rembg", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/danielgatis/rembg/HEAD/LICENSE.txt", "family": "permissive", "id": "MIT"}, "name": "rembg", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/xinntao/Real-ESRGAN", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "image upscaling", "description": "Enlarges and cleans up images and old photographs.", "kind": "vision", "id": "u-realesrgan", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/xinntao/Real-ESRGAN/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "Real-ESRGAN", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/BAAI/bge-m3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding model", "description": "Turns text into vectors across 100+ languages, handling long passages, and does dense and sparse retrieval at once.", "kind": "vectors", "id": "u-bgem3", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "BGE-M3", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/nomic-ai/nomic-embed-text-v1.5", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding model", "description": "A text embedding model with a long context window and open training data.", "kind": "vectors", "id": "u-nomicembed", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "nomic-embed-text", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/intfloat/multilingual-e5-large", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding model", "description": "Multilingual text embeddings, in several sizes so you can trade quality for speed.", "kind": "vectors", "id": "u-e5", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "multilingual-e5", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/jinaai/jina-embeddings-v3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding model", "description": "Multilingual embeddings with task-specific adapters and long input support.", "kind": "vectors", "id": "u-jinaembed", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "jina-embeddings-v3", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/Alibaba-NLP/gte-multilingual-base", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding model", "description": "General text embeddings from Alibaba, in several sizes.", "kind": "vectors", "id": "u-gte", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "GTE", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://huggingface.co/BAAI/bge-reranker-v2-m3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["score"], "inputs": ["text"]}, "what_it_is": "reranker", "description": "Takes the results a search returned and reorders them by how well they really answer the question. Usually the cheapest large quality gain in a RAG system.", "kind": "vectors", "id": "u-bgereranker", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "BGE Reranker", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/sentence-transformers", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding library", "description": "The usual way to compute, compare and fine-tune text embeddings in Python.", "kind": "vectors", "id": "u-sentencetransformers", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/sentence-transformers/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Sentence Transformers", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/stanford-futuredata/ColBERT", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector", "score"], "inputs": ["text"]}, "what_it_is": "late interaction retrieval", "description": "Retrieval that keeps a vector per token instead of one per document, which finds things a single vector misses.", "kind": "vectors", "id": "u-colbert", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/stanford-futuredata/ColBERT/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "ColBERT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/naver/splade", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "sparse retrieval", "description": "Learned sparse retrieval: keeps the interpretability of keyword search with much of the quality of embeddings.", "kind": "vectors", "id": "u-splade", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "Not for commercial use at all - you cannot sell anything built on it.", "proof": "https://raw.githubusercontent.com/naver/splade/HEAD/LICENSE", "family": "non_commercial", "id": "CC-BY-NC"}, "name": "SPLADE", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/unclecode/crawl4ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON"], "inputs": ["URL"]}, "what_it_is": "web crawler for models", "description": "Crawls web pages and returns clean Markdown made for feeding a model, not a browser.", "kind": "tools", "id": "u-crawl4ai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/unclecode/crawl4ai/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Crawl4AI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/firecrawl/firecrawl", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON"], "inputs": ["URL"]}, "what_it_is": "site to markdown", "description": "Turns a whole site into Markdown or structured data, following links as it goes.", "kind": "tools", "id": "u-firecrawl", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/firecrawl/firecrawl/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Firecrawl", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/playwright", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["HTML", "image"], "inputs": ["URL"]}, "what_it_is": "browser automation", "description": "Drives a real browser from code: clicks, types, waits and takes screenshots.", "kind": "tools", "id": "u-playwright", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/playwright/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Playwright", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "It is a driver. Timing it measures the page you point it at.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it, and we still do: it drives a real browser for us, including the screen recordings we make of our own work."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/browser-use/browser-use", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image"], "inputs": ["URL", "text"]}, "what_it_is": "browser for agents", "description": "Lets an agent use a browser the way a person does, describing pages back to the model.", "kind": "tools", "id": "u-browseruse", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/browser-use/browser-use/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "browser-use", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed; what it costs is model calls, not milliseconds.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it - a browser an agent can drive by itself rather than by selectors written by hand."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/dottxt-ai/outlines", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "schema"]}, "what_it_is": "structured output", "description": "Forces a model's output to follow a JSON schema, a regular expression or a grammar, so it cannot drift out of format.", "kind": "tools", "id": "u-outlines", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/dottxt-ai/outlines/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Outlines", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/567-labs/instructor", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "schema"]}, "what_it_is": "typed model output", "description": "Gets typed, validated objects out of a model instead of text you have to parse.", "kind": "tools", "id": "u-instructor", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/567-labs/instructor/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Instructor", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/guidance-ai/guidance", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "schema"]}, "what_it_is": "constrained generation", "description": "Interleaves your control flow with the model's generation, so structure is enforced as it writes. Made by Microsoft.", "kind": "tools", "id": "u-guidance", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/guidance-ai/guidance/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "Guidance", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/adbar/trafilatura", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "Markdown"], "inputs": ["HTML", "URL"]}, "what_it_is": "text from web pages", "description": "Pulls the actual article out of a web page and throws away the navigation and adverts.", "kind": "tools", "id": "u-trafilatura", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/adbar/trafilatura/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Trafilatura", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mozilla/readability", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["HTML", "text"], "inputs": ["HTML"]}, "what_it_is": "article extraction", "description": "The reading-mode algorithm as a library: page in, article out.", "kind": "tools", "id": "u-readability", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mozilla/readability/HEAD/LICENSE.md", "family": "permissive", "id": "Apache-2.0"}, "name": "Readability", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/searxng/searxng", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "private metasearch", "description": "A search engine you host yourself that queries others and keeps no profile of you.", "kind": "tools", "id": "u-searxng", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/searxng/searxng/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "SearXNG", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deedy5/ddgs", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "web search from code", "description": "Search the web from a script without an API key.", "kind": "tools", "id": "u-duckduckgosearch", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/deedy5/ddgs/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "ddgs", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/e2b-dev/E2B", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "file"], "inputs": ["code"]}, "what_it_is": "sandbox for agent code", "description": "Runs code an agent wrote inside an isolated sandbox, so a mistake stays inside it.", "kind": "tools", "id": "u-e2b", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/e2b-dev/E2B/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "E2B", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/BerriAI/litellm", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "one API for many models", "description": "One OpenAI-shaped interface in front of a hundred model providers, with cost tracking.", "kind": "tools", "id": "u-litellm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/BerriAI/litellm/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LiteLLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "It is a layer in front of someone else's model - timing it would measure them, plus a little of our network.", "label": "TESTED", "machine": "Windows 11, and our Linux server", "verdict": "We ran it. One OpenAI-shaped API in front of many providers, which is the thing that lets you swap a model without touching the code that calls it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/PrefectHQ/fastmcp", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["MCP server"], "inputs": ["code"]}, "what_it_is": "build MCP servers", "description": "A Python framework for writing Model Context Protocol servers and clients.", "kind": "tools", "id": "u-fastmcp", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/PrefectHQ/fastmcp/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "FastMCP", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/EleutherAI/lm-evaluation-harness", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["model"]}, "what_it_is": "model benchmarks", "description": "The harness most open leaderboards run on: dozens of benchmarks, one command.", "kind": "evaluation", "id": "u-lmevalharness", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/EleutherAI/lm-evaluation-harness/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "lm-evaluation-harness", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/promptfoo/promptfoo", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "schema"]}, "what_it_is": "prompt testing", "description": "Tests and compares prompts and models side by side, with assertions that fail a build.", "kind": "evaluation", "id": "u-promptfoo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/promptfoo/promptfoo/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "promptfoo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/vibrantlabsai/ragas", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["score", "JSON"], "inputs": ["text"]}, "what_it_is": "RAG evaluation", "description": "Scores a retrieval system on faithfulness, relevance and context quality.", "kind": "evaluation", "id": "u-ragas", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/vibrantlabsai/ragas/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Ragas", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/confident-ai/deepeval", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["score", "JSON"], "inputs": ["text"]}, "what_it_is": "LLM testing", "description": "Unit tests for model output, run the way you run pytest.", "kind": "evaluation", "id": "u-deepeval", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/confident-ai/deepeval/HEAD/LICENSE.md", "family": "permissive", "id": "Apache-2.0"}, "name": "DeepEval", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/langfuse/langfuse", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["trace"]}, "what_it_is": "tracing and evals", "description": "Records what your agent did, what it cost and how long it took, and keeps the traces to compare against later.", "kind": "evaluation", "id": "u-langfuse", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/langfuse/langfuse/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Langfuse", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Arize-ai/phoenix", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["trace"]}, "what_it_is": "LLM observability", "description": "Open tracing and evaluation for model applications, with a UI you host yourself.", "kind": "evaluation", "id": "u-phoenix", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/Arize-ai/phoenix/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Phoenix", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/guardrails-ai/guardrails", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "schema"]}, "what_it_is": "output validation", "description": "Checks a model's output against rules and re-asks when it fails.", "kind": "evaluation", "id": "u-guardrails", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/guardrails-ai/guardrails/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Guardrails", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/NVIDIA/garak", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["model"]}, "what_it_is": "LLM vulnerability scanner", "description": "Probes a model for prompt injection, leakage and jailbreaks, the way a scanner probes a server.", "kind": "evaluation", "id": "u-garak", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/NVIDIA/garak/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "garak", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/unslothai/unsloth", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "fast fine-tuning", "description": "Fine-tunes models noticeably faster and in less memory, so a single consumer card is often enough.", "kind": "training", "id": "u-unsloth", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/unslothai/unsloth/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Unsloth", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/axolotl-ai-cloud/axolotl", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "fine-tuning framework", "description": "Fine-tuning driven by a config file, covering most of the usual methods.", "kind": "training", "id": "u-axolotl", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Axolotl", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "A training run's time depends on the dataset and the settings, so a single number here would say nothing about yours.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, for fine-tuning."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/peft", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "parameter-efficient tuning", "description": "LoRA and friends: trains a small number of new weights instead of the whole model.", "kind": "training", "id": "u-peft", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/peft/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "PEFT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/trl", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "preference training", "description": "Trains models on human or AI preferences, with the usual algorithms implemented.", "kind": "training", "id": "u-trl", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/trl/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "TRL", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/hiyouga/LlamaFactory", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "training UI", "description": "Fine-tunes a hundred model families from a web interface or the command line.", "kind": "training", "id": "u-llamafactory", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/hiyouga/LlamaFactory/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "LLaMA-Factory", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/casper-hansen/AutoAWQ", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model"]}, "what_it_is": "quantization", "description": "Shrinks a model to 4 bits while keeping most of its quality, for faster inference.", "kind": "training", "id": "u-autoawq", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/casper-hansen/AutoAWQ/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "AutoAWQ", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ModelCloud/GPTQModel", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model"]}, "what_it_is": "quantization", "description": "GPTQ quantization, kept up to date with current model architectures.", "kind": "training", "id": "u-gptqmodel", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ModelCloud/GPTQModel/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "GPTQModel", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "How long a quantisation takes depends on the model going in, so a single figure would say nothing about yours.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, quantising models ourselves. Quantisation is the step that decides whether a model fits the card at all, and our own numbers on this shelf are almost all taken at q4 for exactly that reason."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/vllm-project/llm-compressor", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model"]}, "what_it_is": "quantization and sparsity", "description": "Compresses models for vLLM: quantization, pruning and sparsity in one tool.", "kind": "training", "id": "u-llmcompressor", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/vllm-project/llm-compressor/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "LLM Compressor", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed, same reason.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Same job as the other quantiser here, with sparsity on top."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mostlygeek/llama-swap", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "model switcher", "description": "Sits in front of llama.cpp and swaps models on demand, so one machine can serve several.", "kind": "run locally", "id": "u-llamaswap", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mostlygeek/llama-swap/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "llama-swap", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "The time it saves is the load time of whatever model you swap to.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. It swaps which model llama.cpp is holding without restarting the server - worth knowing, because a cold start of a large quantised model is the kind of wait we have measured in minutes, not seconds."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/LostRuins/koboldcpp", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image"], "inputs": ["text", "image"]}, "what_it_is": "single-file runner", "description": "One executable that runs GGUF models with a web interface, image generation and speech built in. Runs entirely offline on your own machine — no account, and it does not send your data to the cloud or anywhere else.", "kind": "run locally", "id": "u-koboldcpp", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/LostRuins/koboldcpp/HEAD/LICENSE.md", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "KoboldCpp", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/oobabooga/textgen", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "local model UI", "description": "A web interface for running local models, with several backends and an extension system. Everything runs locally: private by construction, and it keeps working with the internet unplugged.", "kind": "run locally", "id": "u-textgenwebui", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/oobabooga/textgen/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Text Generation WebUI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mudler/LocalAI", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image", "audio"], "inputs": ["text", "image", "audio"]}, "what_it_is": "OpenAI-compatible server", "description": "A drop-in replacement for the OpenAI API that runs models on your own hardware, including images and audio. Runs offline, on your own hardware: no internet connection needed once the model is downloaded, and it does not send your data to the cloud.", "kind": "run locally", "id": "u-localai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mudler/LocalAI/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LocalAI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/sgl-project/sglang", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "serving engine", "description": "A serving engine built for structured output and heavy concurrency.", "kind": "run locally", "id": "u-sglang", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/sgl-project/sglang/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "SGLang", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/text-generation-inference", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "serving engine", "description": "A production server for text generation, with batching and quantization. Made by Hugging Face.", "kind": "run locally", "id": "u-tgi", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/text-generation-inference/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Text Generation Inference", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ml-explore/mlx", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image"], "inputs": ["text", "image"]}, "what_it_is": "Apple silicon framework", "description": "A machine-learning framework built for Apple silicon's shared memory.", "kind": "run locally", "id": "u-mlx", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ml-explore/mlx/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MLX", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenNMT/CTranslate2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "audio"]}, "what_it_is": "fast inference engine", "description": "An inference engine for transformer models that is markedly faster on CPU. It is what faster-whisper runs on.", "kind": "run locally", "id": "u-ctranslate2", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenNMT/CTranslate2/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "CTranslate2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/janhq/jan", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "desktop AI app", "description": "A desktop application that runs models locally and can also talk to remote ones. Works offline with local models, so your conversations stay on your own machine instead of going to a cloud service.", "kind": "run locally", "id": "u-jan", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/janhq/jan/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Jan", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/open-webui/open-webui", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "file"]}, "what_it_is": "self-hosted chat UI", "description": "A chat interface you host yourself, for local or remote models, with users and documents. Self-hosted: with a local model behind it, the whole chat is private and works offline.", "kind": "run locally", "id": "u-openwebui", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/open-webui/open-webui/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "Open WebUI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "It is an interface. Its speed is the speed of the model behind it.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11, and our Linux server", "verdict": "We ran it. A chat interface you host yourself in front of your own model, which is the piece that turns a local model into something you can hand to someone else without teaching them a command line."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/danny-avila/LibreChat", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "file"]}, "what_it_is": "self-hosted chat", "description": "A chat platform you run yourself that talks to many model providers at once. Self-hosted, so with local models it works offline and keeps conversations on your own server.", "kind": "run locally", "id": "u-librechat", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/danny-avila/LibreChat/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LibreChat", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/gpustack/gpustack", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "GPU cluster manager", "description": "Pools several machines' graphics cards into one place to serve models from. Your own machines, your own models: private AI without a cloud provider.", "kind": "run locally", "id": "u-gpustack", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/gpustack/gpustack/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "GPUStack", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/containers/ramalama", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "models in containers", "description": "Runs models as OCI containers, so a model is handled like any other piece of software. The model runs on your own hardware, offline, with nothing sent to a cloud provider.", "kind": "run locally", "id": "u-ramalama", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/containers/ramalama/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "RamaLama", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/autogen", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "multi-agent framework", "description": "Builds systems where several agents talk to each other to finish a task. Made by Microsoft.", "kind": "agents", "id": "u-autogen", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "Usable commercially, but you must credit the author visibly.", "proof": "https://raw.githubusercontent.com/microsoft/autogen/HEAD/LICENSE", "family": "permissive_with_credit", "id": "CC-BY-4.0"}, "name": "AutoGen", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/run-llama/llama_index", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "file"]}, "what_it_is": "data framework", "description": "Connects your documents and databases to a model: indexing, retrieval and agents.", "kind": "agents", "id": "u-llamaindex", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/run-llama/llama_index/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LlamaIndex", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/stanfordnlp/dspy", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "programming, not prompting", "description": "You declare what you want and it compiles the prompts, optimising them against examples.", "kind": "agents", "id": "u-dspy", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/stanfordnlp/dspy/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "DSPy", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/smolagents", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "code"], "inputs": ["text"]}, "what_it_is": "small agent library", "description": "A very small agent library where the agent writes Python to act, instead of emitting JSON.", "kind": "agents", "id": "u-smolagents", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/smolagents/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "smolagents", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pydantic/pydantic-ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "schema"]}, "what_it_is": "typed agents", "description": "An agent framework built around typed, validated results. From the Pydantic team.", "kind": "agents", "id": "u-pydanticai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/pydantic/pydantic-ai/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Pydantic AI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenHands/OpenHands", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["text", "code"]}, "what_it_is": "software agent", "description": "An agent that writes and runs code, uses a browser and works through development tasks.", "kind": "agents", "id": "u-openhands", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenHands/OpenHands/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "OpenHands", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. An agent that writes and runs code on its own; what it costs is model calls, and that is the number to watch, not seconds."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/agno-agi/agno", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "agent runtime", "description": "A runtime for building agents with memory, tools and knowledge, aimed at low overhead.", "kind": "agents", "id": "u-agno", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/agno-agi/agno/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Agno", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/langflow-ai/langflow", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "visual flow builder", "description": "Builds agent and retrieval flows by dragging boxes, then exports them as an API.", "kind": "agents", "id": "u-langflow", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/langflow-ai/langflow/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Langflow", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/activepieces/activepieces", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "automation platform", "description": "An open automation platform with AI steps, in the shape of the commercial ones.", "kind": "agents", "id": "u-activepieces", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/activepieces/activepieces/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Activepieces", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/windmill-labs/windmill", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["code"]}, "what_it_is": "scripts and workflows", "description": "Turns scripts into workflows and internal tools, with schedules and permissions.", "kind": "agents", "id": "u-windmill", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/windmill-labs/windmill/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Windmill", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/temporalio/temporal", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "durable execution", "description": "Runs long workflows that survive crashes and restarts, which is what agent runs need once they take hours.", "kind": "agents", "id": "u-temporal", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/temporalio/temporal/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Temporal", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/facebookresearch/faiss", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["vector"]}, "what_it_is": "vector search library", "description": "The library most vector databases are built on. Searches billions of vectors. Made by Meta.", "kind": "memory", "id": "u-faiss", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/facebookresearch/faiss/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "FAISS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/unum-cloud/usearch", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["vector"]}, "what_it_is": "compact vector search", "description": "A very small and fast vector search engine that embeds into almost any language.", "kind": "memory", "id": "u-usearch", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/unum-cloud/usearch/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "USearch", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/neuml/txtai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "vector"]}, "what_it_is": "embeddings database", "description": "An embeddings database with search, retrieval pipelines and workflows in one package.", "kind": "memory", "id": "u-txtai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/neuml/txtai/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "txtai", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/marqo-ai/marqo", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "image"]}, "what_it_is": "end-to-end vector search", "description": "Vector search that does the embedding for you, for text and images.", "kind": "memory", "id": "u-marqo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/marqo-ai/marqo/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Marqo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/typesense/typesense", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "vector"]}, "what_it_is": "search engine", "description": "A fast typo-tolerant search engine that also does vector search.", "kind": "memory", "id": "u-typesense", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/typesense/typesense/HEAD/LICENSE.txt", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "Typesense", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/meilisearch/meilisearch", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "vector"]}, "what_it_is": "search engine", "description": "A search engine that is quick to set up and has hybrid keyword and vector search.", "kind": "memory", "id": "u-meilisearch", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/meilisearch/meilisearch/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Meilisearch", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/opensearch-project/OpenSearch", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "vector"]}, "what_it_is": "search and analytics", "description": "A search and analytics engine with a vector plugin, forked from Elasticsearch.", "kind": "memory", "id": "u-opensearch", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/opensearch-project/OpenSearch/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "OpenSearch", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11, and our Linux server", "verdict": "We ran it, as a search and analytics store."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mem0ai/mem0", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "memory layer for agents", "description": "Gives an agent memory that persists across conversations, deciding what is worth keeping.", "kind": "memory", "id": "u-mem0", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mem0ai/mem0/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Mem0", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/graphrag", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "graph"], "inputs": ["text"]}, "what_it_is": "graph-based retrieval", "description": "Builds a knowledge graph from your documents so questions that span many of them can be answered. Made by Microsoft.", "kind": "memory", "id": "u-graphrag", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/graphrag/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "GraphRAG", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/HKUDS/LightRAG", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "graph"], "inputs": ["text"]}, "what_it_is": "lighter graph retrieval", "description": "Graph-plus-vector retrieval aimed at being much cheaper to build and update than a full knowledge graph.", "kind": "memory", "id": "u-lightrag", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/HKUDS/LightRAG/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LightRAG", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/infiniflow/ragflow", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["PDF", "DOCX", "text"]}, "what_it_is": "retrieval with deep parsing", "description": "A retrieval engine built around understanding documents properly before chunking them.", "kind": "memory", "id": "u-ragflow", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/infiniflow/ragflow/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "RAGFlow", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/black-forest-labs/flux", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text", "image"]}, "what_it_is": "image model", "description": "An image generation model family with a well-regarded open variant.", "kind": "image", "id": "u-flux", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/black-forest-labs/flux/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "FLUX.1", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. This is the image model our own work settled on — the numbers we published for image generation were taken on the schnell build at q4."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/AUTOMATIC1111/stable-diffusion-webui", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text", "image"]}, "what_it_is": "image UI", "description": "The long-standing web interface for Stable Diffusion, with a very large extension ecosystem.", "kind": "image", "id": "u-sdwebui", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/HEAD/LICENSE.txt", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Stable Diffusion WebUI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Worth knowing before you build a product on it: it is AGPL, which is why the licence badge sits on this card."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/lllyasviel/stable-diffusion-webui-forge", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text", "image"]}, "what_it_is": "faster image UI", "description": "A fork of the Stable Diffusion interface focused on speed and lower memory use.", "kind": "image", "id": "u-forge", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/lllyasviel/stable-diffusion-webui-forge/HEAD/LICENSE.txt", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "SD WebUI Forge", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Faster than the interface it forked from, and it carries the same AGPL terms — the badge on this card is not decoration."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/invoke-ai/InvokeAI", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text", "image"]}, "what_it_is": "image studio", "description": "An image generation studio with a canvas, layers and model management.", "kind": "image", "id": "u-invokeai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/invoke-ai/InvokeAI/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "InvokeAI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/lllyasviel/Fooocus", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text"]}, "what_it_is": "simple image generation", "description": "Image generation with almost no settings: you type, it produces something good.", "kind": "image", "id": "u-fooocus", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/lllyasviel/Fooocus/HEAD/LICENSE", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "Fooocus", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/lllyasviel/ControlNet", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image", "text"]}, "what_it_is": "guided image generation", "description": "Controls an image model with a pose, an edge map or a depth map, instead of only words.", "kind": "image", "id": "u-controlnet", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/lllyasviel/ControlNet/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "ControlNet", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/diffusers", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image", "video"], "inputs": ["text", "image"]}, "what_it_is": "diffusion library", "description": "The library most image and video models are shipped through.", "kind": "image", "id": "u-diffusers", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/diffusers/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Diffusers", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. It is the library underneath most of the image work here, rather than something you open."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/vladmandic/sdnext", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text", "image"]}, "what_it_is": "image generation platform", "description": "An image generation interface that supports a wide spread of model architectures.", "kind": "image", "id": "u-sdnext", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/vladmandic/sdnext/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "SD.Next", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bmaltais/kohya_ss", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["image", "text"]}, "what_it_is": "image model training", "description": "Trains LoRAs and fine-tunes image models from a graphical interface.", "kind": "image", "id": "u-kohya", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/bmaltais/kohya_ss/HEAD/LICENSE.md", "family": "permissive", "id": "Apache-2.0"}, "name": "Kohya GUI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ace-step/ACE-Step", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "music generation", "description": "Generates full songs from a text description and lyrics, including the singing voice. Runs on a single consumer card.", "kind": "sound", "id": "u-acestep", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ace-step/ACE-Step/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "ACE-Step", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$about_the_machine": "This number is about the machine named in `machine`, not about the tool in general. A different card gives a different number, and we do not have it.", "details": {"with_the_server_already_running": "~52 s per track", "output_format": "48 kHz WAV, RMS -15 dB, 0% clipping", "cold_start": "~97 s per track (the language model loads once, ~80 s)", "romana": "detected automatically; the Romanian accent is right on the turbo model", "varianta_XL_4B": "fits in 12 GB but takes 114 s for a short track and 213 s for two minutes — and its Romanian accent is noticeably worse than the small turbo model", "varianta_base_4B": "~91 s per track"}, "what_we_measured": "Romanian song generation, turbo (2B), on the GPU", "figure": "38 s to generate a track", "how": "acestep-v15-turbo DiT + acestep-5Hz-lm-1.7B, offload_to_cpu=True, inference_steps=8, shift=3.0. Quality judged by ear by the person who speaks the language, not by a score.", "date": "2026-07-04", "worth_knowing": "The server holds about 11 GB of 12 GB EVEN WHEN IDLE — offload_to_cpu does not give VRAM back. If you want the card for anything else, stop the server after each track; VRAM drops from ~11 GB to ~1 GB. Also: the turbo model does not use CFG, so negative prompts are silently ignored on it.", "label": "MEASURED", "the_whole_chain": ["ACE-Step", "FFmpeg"], "machine": "RTX 5070 12GB, Windows 11, CUDA"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/yt-dlp/yt-dlp", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video", "audio"], "inputs": ["URL"]}, "what_it_is": "media downloader", "description": "Downloads video and audio from a very large number of sites. The first step of most transcription and subtitling chains.", "kind": "video", "id": "u-ytdlp", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/yt-dlp/yt-dlp/HEAD/LICENSE", "family": "permissive", "id": "Unlicense"}, "name": "yt-dlp", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-09-14", "why_no_figure": "Its time is network time, which says more about the connection than about the tool.", "label": "TESTED", "machine": "Windows 11", "verdict": "Ran it as the first step of our speech-to-text chain, without a single failure."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/guoyww/AnimateDiff", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["text", "image"]}, "what_it_is": "animation from image models", "description": "Turns a still image model into an animation model without retraining it.", "kind": "video", "id": "u-animatediff", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/guoyww/AnimateDiff/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "AnimateDiff", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Our own video work went to LTX-Video instead, so this did not become part of the pipeline."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/lllyasviel/FramePack", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["image", "text"]}, "what_it_is": "long video on small cards", "description": "Generates longer video on modest hardware by packing frame context.", "kind": "video", "id": "u-framepack", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/lllyasviel/FramePack/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "FramePack", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/FFmpeg/FFmpeg", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video", "audio", "image"], "inputs": ["video", "audio", "image"]}, "what_it_is": "audio and video toolkit", "description": "Converts, cuts, joins, re-encodes and streams almost any audio or video format. Nearly every media tool on this shelf calls it underneath.", "kind": "video", "id": "u-ffmpeg", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can link it into a closed product, but changes to THIS library must stay open.", "proof": "https://raw.githubusercontent.com/FFmpeg/FFmpeg/HEAD/LICENSE.md", "family": "weak_copyleft", "id": "LGPL-2.1"}, "name": "FFmpeg", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-09-14", "why_no_figure": "The conversion is a rounding error next to the transcription, so a number here would suggest a choice that does not exist.", "label": "TESTED", "machine": "Windows 11", "verdict": "Ran it to convert phone audio to 16 kHz mono WAV. Nothing surprising, which for this job is the highest praise."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Zulko/moviepy", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["video", "audio", "image"]}, "what_it_is": "video editing from code", "description": "Cuts, concatenates and composites video from a Python script.", "kind": "video", "id": "u-moviepy", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "MoviePy", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/WyattBlue/auto-editor", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video", "audio"], "inputs": ["video", "audio"]}, "what_it_is": "automatic cutting", "description": "Removes silence and unwanted sections from a recording automatically.", "kind": "video", "id": "u-autoeditor", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/WyattBlue/auto-editor/HEAD/LICENSE", "family": "permissive", "id": "Unlicense"}, "name": "Auto-Editor", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/hzwer/ECCV2022-RIFE", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["video"]}, "what_it_is": "frame interpolation", "description": "Invents the frames between frames, to smooth or slow down video.", "kind": "video", "id": "u-rife", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/hzwer/ECCV2022-RIFE/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "RIFE", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/datasets", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["dataset"]}, "what_it_is": "dataset library", "description": "Loads and processes datasets of any size without holding them in memory.", "kind": "data", "id": "u-datasets", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/datasets/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Datasets", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/duckdb/duckdb", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["CSV", "Parquet", "SQL"]}, "what_it_is": "analytical database", "description": "A database in a single file that runs analytical SQL over CSV and Parquet at speed.", "kind": "data", "id": "u-duckdb", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/duckdb/duckdb/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "DuckDB", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it, as the analytical store for our own measurement data."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pola-rs/polars", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["CSV", "Parquet"]}, "what_it_is": "dataframes", "description": "A dataframe library that is much faster than the usual one on large files.", "kind": "data", "id": "u-polars", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/pola-rs/polars/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Polars", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/datatrove", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["text"]}, "what_it_is": "text pipeline at scale", "description": "Filters and deduplicates very large text corpora for training.", "kind": "data", "id": "u-datatrove", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/datatrove/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "DataTrove", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/argilla-io/distilabel", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["text"]}, "what_it_is": "synthetic data", "description": "Builds synthetic datasets with models, with the generation steps written down.", "kind": "data", "id": "u-distilabel", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/argilla-io/distilabel/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "distilabel", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/argilla-io/argilla", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["text"]}, "what_it_is": "data annotation", "description": "A tool for humans to label and curate the data a model will learn from.", "kind": "data", "id": "u-argilla", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/argilla-io/argilla/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Argilla", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/HumanSignal/label-studio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["text", "image", "audio", "video"]}, "what_it_is": "labelling for any data", "description": "Labels text, images, audio and video in one interface.", "kind": "data", "id": "u-labelstudio", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/HumanSignal/label-studio/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Label Studio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it, for labelling data by hand."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/fivetran/great_expectations", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["table"]}, "what_it_is": "data validation", "description": "Writes down what your data must look like, and fails loudly when it stops looking like that.", "kind": "data", "id": "u-greatexpectations", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/fivetran/great_expectations/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Great Expectations", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/data-privacy-stack/presidio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "image"]}, "what_it_is": "personal data detection", "description": "Finds and hides personal data in text and images.", "kind": "evaluation", "id": "u-presidio", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/data-privacy-stack/presidio/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Presidio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/protectai/rebuff", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "prompt injection defence", "description": "Detects prompt injection attempts in several layers.", "kind": "evaluation", "id": "u-rebuff", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/protectai/rebuff/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Rebuff", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/protectai/modelscan", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["model"]}, "what_it_is": "model file scanning", "description": "Scans model files for code that would run when the file is loaded.", "kind": "evaluation", "id": "u-modelscan", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/protectai/modelscan/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "ModelScan", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/NVIDIA-NeMo/Guardrails", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "conversation rails", "description": "Puts programmable rails on what a conversational system may and may not do. Made by NVIDIA.", "kind": "evaluation", "id": "u-nemoguardrails", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/NVIDIA-NeMo/Guardrails/HEAD/LICENSE.md", "family": "custom", "id": "custom"}, "name": "NeMo Guardrails", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/QwenLM/Qwen3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "A family of open models from small to very large, strong on many languages and on code.", "kind": "models", "id": "u-qwen3", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Qwen3", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran this family more than any other: the brain behind our own work has been a Qwen for months, and most measured numbers on this shelf were taken on one."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/meta-llama/llama-models", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "Meta's open model family, the most widely fine-tuned base in the world.", "kind": "models", "id": "u-llamamodels", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/meta-llama/llama-models/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Llama", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran models from this family as candidate brains."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mistralai/mistral-inference", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "Compact open models known for punching above their size.", "kind": "models", "id": "u-mistralmodels", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mistralai/mistral-inference/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Mistral models", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-06", "why_no_figure": "We judged it on the quality of its Romanian, not on a stopwatch, and that is the honest thing to report about it.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran RoMistral-7B-Instruct (Q4_K_M) as a candidate brain. Its Romanian was genuinely good — the reason we kept it around for a while — and we later dropped it for a larger mixture-of-experts model, not because it was bad."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/google-deepmind/gemma", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "Google's open model family, including small variants that fit modest cards.", "kind": "models", "id": "u-gemma", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/google-deepmind/gemma/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Gemma", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran models from this family as candidate brains."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deepseek-ai/DeepSeek-V3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "Open models with strong reasoning and coding results for their size.", "kind": "models", "id": "u-deepseek", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "DeepSeek", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran models from this family."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/allenai/OLMo", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "fully open model", "description": "A model where the data, the code and the training logs are published, not only the weights.", "kind": "models", "id": "u-olmo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/allenai/OLMo/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "OLMo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Fully open — weights, data and training code — which is rarer than the word 'open' usually means on a model page."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/smollm", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "tiny models", "description": "Very small models meant to run on a phone or in a browser.", "kind": "models", "id": "u-smollm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/smollm/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "SmolLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. Small enough to run where nothing else fits, and we measured how fast that gets: our tiny-model numbers are on this shelf."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bigcode-project/starcoder2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "code model", "description": "A code model trained on a documented, permissively licensed corpus.", "kind": "models", "id": "u-starcoder", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/bigcode-project/starcoder2/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "StarCoder2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, as a code model."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/huggingface_hub", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "dataset"], "inputs": ["text"]}, "what_it_is": "model downloads", "description": "Downloads, uploads and manages models and datasets from the command line.", "kind": "models", "id": "u-hfhub", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/huggingface_hub/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Hugging Face Hub CLI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/modelcontextprotocol/servers", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["MCP server"], "inputs": ["text"]}, "what_it_is": "example MCP servers", "description": "The reference servers for the Model Context Protocol: files, git, fetch and more.", "kind": "tools", "id": "u-mcpservers", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/modelcontextprotocol/servers/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MCP reference servers", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/modelcontextprotocol/python-sdk", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["MCP server"], "inputs": ["code"]}, "what_it_is": "build MCP in Python", "description": "The official Python SDK for writing Model Context Protocol servers and clients.", "kind": "tools", "id": "u-mcppython", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/modelcontextprotocol/python-sdk/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MCP Python SDK", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/modelcontextprotocol/typescript-sdk", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["MCP server"], "inputs": ["code"]}, "what_it_is": "build MCP in TypeScript", "description": "The official TypeScript SDK for the Model Context Protocol.", "kind": "tools", "id": "u-mcpts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/modelcontextprotocol/typescript-sdk/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MCP TypeScript SDK", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openai/tiktoken", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["count"], "inputs": ["text"]}, "what_it_is": "token counting", "description": "Counts tokens the way the model does, instead of guessing from characters.", "kind": "tools", "id": "u-tiktoken", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openai/tiktoken/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "tiktoken", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/gradio-app/gradio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["web app"], "inputs": ["code"]}, "what_it_is": "quick UI for models", "description": "Wraps a model in a web interface in a few lines, for demos and internal tools.", "kind": "tools", "id": "u-gradio", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/gradio-app/gradio/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Gradio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/streamlit/streamlit", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["web app"], "inputs": ["code"]}, "what_it_is": "data apps", "description": "Turns a Python script into a web application without front-end work.", "kind": "tools", "id": "u-streamlit", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/streamlit/streamlit/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Streamlit", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Chainlit/chainlit", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["web app"], "inputs": ["code"]}, "what_it_is": "chat UI for agents", "description": "A chat interface for an agent that shows the intermediate steps it took.", "kind": "tools", "id": "u-chainlit", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Chainlit/chainlit/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Chainlit", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/vercel/ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "AI in TypeScript apps", "description": "A TypeScript toolkit for streaming model output into web applications.", "kind": "tools", "id": "u-vercelai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/vercel/ai/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Vercel AI SDK", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it, building against it in TypeScript."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/transformers.js", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "image"]}, "what_it_is": "models in the browser", "description": "Runs models directly in the browser with WebGPU, with nothing sent to a server.", "kind": "tools", "id": "u-transformersjs", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/transformers.js/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Transformers.js", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/onnxruntime", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "vector"], "inputs": ["model"]}, "what_it_is": "cross-platform inference", "description": "Runs models exported to ONNX on almost any hardware and language.", "kind": "tools", "id": "u-onnxruntime", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/onnxruntime/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "ONNX Runtime", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openvinotoolkit/openvino", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image"], "inputs": ["model"]}, "what_it_is": "inference on Intel hardware", "description": "Optimises and runs models on Intel processors, graphics and NPUs.", "kind": "tools", "id": "u-openvino", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openvinotoolkit/openvino/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "OpenVINO", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bentoml/BentoML", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["web app"], "inputs": ["model"]}, "what_it_is": "model serving", "description": "Packages a model and its code into a service you can deploy anywhere.", "kind": "tools", "id": "u-bentoml", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/bentoml/BentoML/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "BentoML", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/transformers", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "vector"], "inputs": ["text", "image", "audio"]}, "what_it_is": "the model library", "description": "The library almost every open model ships with, for running and training them.", "kind": "tools", "id": "u-transformers", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/transformers/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Transformers", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Tencent-Hunyuan/Hunyuan3D-2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["image", "text"]}, "what_it_is": "image to 3D model", "description": "Turns a single image into a textured 3D mesh. Made by Tencent.", "kind": "tools", "id": "u-hunyuan3d", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/Tencent-Hunyuan/Hunyuan3D-2/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Hunyuan3D", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, turning images into 3D models."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/VAST-AI-Research/TripoSR", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["image"]}, "what_it_is": "fast image to 3D", "description": "Reconstructs a 3D mesh from one image in under a second on a decent card.", "kind": "tools", "id": "u-triposr", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/VAST-AI-Research/TripoSR/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "TripoSR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/TencentARC/InstantMesh", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["image"]}, "what_it_is": "single image to mesh", "description": "Makes a 3D mesh from a single image, going through multi-view generation first.", "kind": "tools", "id": "u-instantmesh", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/TencentARC/InstantMesh/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "InstantMesh", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/TRELLIS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["text", "image"]}, "what_it_is": "3D asset generation", "description": "Generates 3D assets from text or images, with several output representations. Made by Microsoft.", "kind": "tools", "id": "u-trellis", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/TRELLIS/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "TRELLIS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We did NOT run this. We weighed it on paper — licence, architecture, how much it occupies, what encoder it drags along — and `why` is the reason that decision came out the way it did. It is not a review of how well it works, because we did not find that out.", "when": "2026-09-18", "why": "Weighed on paper: architecture, size at our quantisation, and whether it fits the card we target. Never started.", "label": "REVIEWED"}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/blender/blender", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh", "image", "video"], "inputs": ["3D mesh", "code"]}, "what_it_is": "3D suite", "description": "The open 3D program: modelling, rig and rigging, animation, rendering. Scriptable in Python, and it exports to game engines, which is what makes it usable from an agent.", "kind": "tools", "id": "u-blender", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/blender/blender/HEAD/COPYING", "family": "custom", "id": "custom"}, "name": "Blender", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. It is the answer on this shelf to 3D and rigging, and it is on the shelf because an agent asked for exactly that and we had nothing."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/alicevision/Meshroom", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["image"]}, "what_it_is": "photogrammetry", "description": "Builds a 3D model out of a set of photographs.", "kind": "tools", "id": "u-meshroom", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Meshroom", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/colmap/colmap", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["point cloud"], "inputs": ["image"]}, "what_it_is": "structure from motion", "description": "Works out camera positions and a point cloud from photographs. The step most 3D reconstruction pipelines start with.", "kind": "tools", "id": "u-colmap", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/colmap/colmap/HEAD/LICENSE", "family": "permissive", "id": "BSD-2-Clause"}, "name": "COLMAP", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/nerfstudio-project/nerfstudio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D scene"], "inputs": ["image", "video"]}, "what_it_is": "neural rendering", "description": "A framework for training and viewing neural radiance fields, with several methods in one place.", "kind": "tools", "id": "u-nerfstudio", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/nerfstudio-project/nerfstudio/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Nerfstudio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/graphdeco-inria/gaussian-splatting", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D scene"], "inputs": ["image", "video"]}, "what_it_is": "real-time scene rendering", "description": "Reconstructs a scene from photographs and renders it in real time.", "kind": "tools", "id": "u-gaussiansplatting", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/graphdeco-inria/gaussian-splatting/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "3D Gaussian Splatting", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/PixarAnimationStudios/OpenUSD", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D scene"], "inputs": ["3D mesh"]}, "what_it_is": "3D scene interchange", "description": "The scene description format that game engines and 3D tools exchange assets through.", "kind": "tools", "id": "u-openusd", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/PixarAnimationStudios/OpenUSD/HEAD/LICENSE.txt", "family": "custom", "id": "custom"}, "name": "OpenUSD", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mikedh/trimesh", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["3D mesh"]}, "what_it_is": "mesh library", "description": "Loads, repairs, measures and converts 3D meshes from Python.", "kind": "tools", "id": "u-trimesh", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mikedh/trimesh/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "Trimesh", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/DepthAnything/Depth-Anything-V2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["depth map"], "inputs": ["image"]}, "what_it_is": "depth from one image", "description": "Estimates depth from a single photograph, which is what turns a flat image into something 3D tools can use.", "kind": "vision", "id": "u-depthanything", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/DepthAnything/Depth-Anything-V2/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Depth Anything V2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/isl-org/MiDaS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["depth map"], "inputs": ["image"]}, "what_it_is": "monocular depth", "description": "The long-standing depth estimation model, trained across many datasets.", "kind": "vision", "id": "u-midas", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/isl-org/MiDaS/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MiDaS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/naver/dust3r", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["point cloud"], "inputs": ["image"]}, "what_it_is": "3D from image pairs", "description": "Reconstructs geometry from two or more images without knowing the camera beforehand.", "kind": "vision", "id": "u-dust3r", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "Not for commercial use at all - you cannot sell anything built on it.", "proof": "https://raw.githubusercontent.com/naver/dust3r/HEAD/LICENSE", "family": "non_commercial", "id": "CC-BY-NC"}, "name": "DUSt3R", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/CMU-Perceptual-Computing-Lab/openpose", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image", "video"]}, "what_it_is": "body pose detection", "description": "Detects body, hand and face keypoints in images and video. The usual first step before rigging a character.", "kind": "vision", "id": "u-openpose", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/CMU-Perceptual-Computing-Lab/openpose/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "OpenPose", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/google-ai-edge/mediapipe", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image", "video"]}, "what_it_is": "on-device perception", "description": "Ready-made pipelines for hands, face, pose and segmentation that run on phones and laptops. Made by Google. On-device and offline, including on Android and iPhone — the camera frames never leave the phone.", "kind": "vision", "id": "u-mediapipe", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/google-ai-edge/mediapipe/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MediaPipe", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Aider-AI/aider", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "pair programming in the terminal", "description": "Edits your repository from the command line, commits as it goes, and works with most models.", "kind": "tools", "id": "u-aider", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Aider-AI/aider/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "Aider", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/cline/cline", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "coding agent in the editor", "description": "An agent inside VS Code that reads, writes and runs, asking before each step.", "kind": "tools", "id": "u-cline", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/cline/cline/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Cline", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/continuedev/continue", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "coding assistant", "description": "An open assistant for VS Code and JetBrains that you point at your own models.", "kind": "tools", "id": "u-continue", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/continuedev/continue/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Continue", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/TabbyML/tabby", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code"]}, "what_it_is": "self-hosted code completion", "description": "Code completion you host yourself, so nothing from your repository leaves the building.", "kind": "tools", "id": "u-tabby", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/TabbyML/tabby/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Tabby", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/SWE-agent/SWE-agent", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "agent that fixes issues", "description": "Takes a GitHub issue and tries to produce the patch that closes it.", "kind": "tools", "id": "u-sweagent", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/SWE-agent/SWE-agent/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "SWE-agent", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/tree-sitter/tree-sitter", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "code parsing", "description": "Parses source code into a syntax tree, fast and incrementally. What most code tools use to understand a file instead of guessing with regular expressions.", "kind": "tools", "id": "u-treesitter", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/tree-sitter/tree-sitter/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Tree-sitter", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/astral-sh/ruff", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code", "JSON"], "inputs": ["code"]}, "what_it_is": "Python linter and formatter", "description": "Checks and formats Python far faster than the tools it replaces.", "kind": "tools", "id": "u-ruff", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/astral-sh/ruff/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Ruff", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/semgrep/semgrep", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "code pattern search", "description": "Finds bugs and unsafe patterns by matching code structure, not text.", "kind": "tools", "id": "u-semgrep", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can link it into a closed product, but changes to THIS library must stay open.", "proof": "https://raw.githubusercontent.com/semgrep/semgrep/HEAD/LICENSE", "family": "weak_copyleft", "id": "LGPL-2.1"}, "name": "Semgrep", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/amazon-science/chronos-forecasting", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["table"]}, "what_it_is": "time series forecasting", "description": "Forecasts time series with a pretrained model, without training one per series. Made by Amazon.", "kind": "data", "id": "u-chronos", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/amazon-science/chronos-forecasting/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Chronos", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/time-series-foundation-models/lag-llama", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["table"]}, "what_it_is": "time series foundation model", "description": "An open foundation model for probabilistic time series forecasting.", "kind": "data", "id": "u-lagllama", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/time-series-foundation-models/lag-llama/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Lag-Llama", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Nixtla/statsforecast", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["table"]}, "what_it_is": "classical forecasting", "description": "Classical statistical forecasting models, implemented fast enough for millions of series.", "kind": "data", "id": "u-statsforecast", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Nixtla/statsforecast/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "StatsForecast", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/unit8co/darts", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["table"]}, "what_it_is": "forecasting library", "description": "One interface over classical and deep forecasting models, so you can compare them fairly.", "kind": "data", "id": "u-darts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/unit8co/darts/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Darts", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/autogluon/autogluon", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "table"], "inputs": ["table", "text", "image"]}, "what_it_is": "automatic machine learning", "description": "Trains strong models on tables, text and images with three lines of code. Made by Amazon.", "kind": "data", "id": "u-autogluon", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/autogluon/autogluon/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "AutoGluon", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/dmlc/xgboost", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "table"], "inputs": ["table"]}, "what_it_is": "gradient boosting", "description": "The gradient boosting library that still wins most competitions on tabular data.", "kind": "data", "id": "u-xgboost", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/dmlc/xgboost/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "XGBoost", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/lightgbm-org/LightGBM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "table"], "inputs": ["table"]}, "what_it_is": "gradient boosting", "description": "Gradient boosting built for speed and large datasets. Made by Microsoft.", "kind": "data", "id": "u-lightgbm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/lightgbm-org/LightGBM/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LightGBM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/catboost/catboost", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "table"], "inputs": ["table"]}, "what_it_is": "gradient boosting", "description": "Gradient boosting that handles categorical columns without you encoding them first.", "kind": "data", "id": "u-catboost", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/catboost/catboost/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "CatBoost", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/PriorLabs/TabPFN", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table"], "inputs": ["table"]}, "what_it_is": "tabular foundation model", "description": "Predicts on small tables in one forward pass, with no training step at all.", "kind": "data", "id": "u-tabpfn", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/PriorLabs/TabPFN/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "TabPFN", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/neo4j/neo4j", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["graph", "JSON"], "inputs": ["graph", "text"]}, "what_it_is": "graph database", "description": "The widely used graph database, with a query language built for relationships.", "kind": "memory", "id": "u-neo4j", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/neo4j/neo4j/HEAD/LICENSE.txt", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "Neo4j", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/memgraph/memgraph", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["graph", "JSON"], "inputs": ["graph"]}, "what_it_is": "in-memory graph database", "description": "A graph database held in memory, for queries that must answer immediately.", "kind": "memory", "id": "u-memgraph", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/memgraph/memgraph/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Memgraph", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/kuzudb/kuzu", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["graph", "JSON"], "inputs": ["graph"]}, "what_it_is": "embedded graph database", "description": "A graph database that runs inside your process, the way SQLite does for tables.", "kind": "memory", "id": "u-kuzu", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/kuzudb/kuzu/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Kuzu", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/topoteretes/cognee", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["graph", "JSON"], "inputs": ["text"]}, "what_it_is": "memory graph for agents", "description": "Builds a graph out of what an agent has seen, so it can be asked about it later.", "kind": "memory", "id": "u-cognee", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/topoteretes/cognee/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "cognee", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/KoljaB/RealtimeSTT", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["audio"]}, "what_it_is": "live transcription", "description": "Transcribes while someone is still speaking, with voice activity detection built in.", "kind": "sound", "id": "u-realtimestt", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/KoljaB/RealtimeSTT/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "RealtimeSTT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/KoljaB/RealtimeTTS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "streaming speech", "description": "Speaks text as it arrives, instead of waiting for the whole sentence.", "kind": "sound", "id": "u-realtimetts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/KoljaB/RealtimeTTS/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "RealtimeTTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/kyutai-labs/moshi", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio", "text"], "inputs": ["audio"]}, "what_it_is": "full-duplex speech model", "description": "A speech model you can interrupt, because it listens and talks at the same time.", "kind": "sound", "id": "u-moshi", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/kyutai-labs/moshi/HEAD/LICENSE-APACHE", "family": "permissive", "id": "Apache-2.0"}, "name": "Moshi", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/parler-tts", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "controllable speech", "description": "Text-to-speech where you describe the voice you want in words.", "kind": "sound", "id": "u-parler", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/parler-tts/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Parler-TTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/espnet/espnet", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio", "text"], "inputs": ["audio", "text"]}, "what_it_is": "speech toolkit", "description": "An end-to-end toolkit covering recognition, synthesis, translation and enhancement.", "kind": "sound", "id": "u-espnet", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/espnet/espnet/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "ESPnet", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/DAMO-NLP-SG/VideoLLaMA3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["video", "text"]}, "what_it_is": "video understanding", "description": "Answers questions about what happens in a video, not just in one frame.", "kind": "video", "id": "u-videollama", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/DAMO-NLP-SG/VideoLLaMA3/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "VideoLLaMA 3", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenGVLab/InternVideo", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["video", "text"]}, "what_it_is": "video foundation models", "description": "A family of models for understanding video: retrieval, captioning, question answering.", "kind": "video", "id": "u-internvideo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenGVLab/InternVideo/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "InternVideo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/dmlc/decord", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["video"]}, "what_it_is": "video reading for models", "description": "Reads video frames efficiently for training and inference, instead of decoding everything.", "kind": "video", "id": "u-decord", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/dmlc/decord/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Decord", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Breakthrough/PySceneDetect", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video", "JSON"], "inputs": ["video"]}, "what_it_is": "shot detection", "description": "Finds where the cuts are in a video and splits it at them.", "kind": "video", "id": "u-scenedetect", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Breakthrough/PySceneDetect/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "PySceneDetect", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/facebookresearch/esm", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh", "vector"], "inputs": ["text"]}, "what_it_is": "protein language models", "description": "Protein models that predict structure and properties from sequence. Made by Meta.", "kind": "models", "id": "u-esm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/facebookresearch/esm/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "ESM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/jwohlwend/boltz", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["3D mesh"], "inputs": ["text"]}, "what_it_is": "biomolecular structure prediction", "description": "Open structure prediction for proteins and their complexes.", "kind": "models", "id": "u-boltz", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/jwohlwend/boltz/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Boltz", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/rdkit/rdkit", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "image"], "inputs": ["text"]}, "what_it_is": "cheminformatics", "description": "The open toolkit for working with molecules in code.", "kind": "models", "id": "u-rdkit", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "RDKit", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deepchem/deepchem", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "table"], "inputs": ["text", "table"]}, "what_it_is": "machine learning for chemistry", "description": "Models and datasets for chemistry, biology and materials science.", "kind": "models", "id": "u-deepchem", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/deepchem/deepchem/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "DeepChem", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/docling-project/docling-serve", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON"], "inputs": ["PDF", "DOCX"]}, "what_it_is": "document parsing as a service", "description": "Runs Docling behind an HTTP API, so an agent can send a document and get structure back.", "kind": "documents", "id": "u-docling-serve", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/docling-project/docling-serve/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Docling Serve", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/agronholm/apscheduler", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "scheduling in Python", "description": "Runs jobs on a schedule inside your own process, which is what an agent needs to check something every hour without a cron file.", "kind": "tools", "id": "u-apscheduler", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/agronholm/apscheduler/HEAD/LICENSE.txt", "family": "permissive", "id": "MIT"}, "name": "APScheduler", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/celery/celery", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "task queue", "description": "Runs work in the background across machines, with retries and scheduling.", "kind": "tools", "id": "u-celery", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/celery/celery/HEAD/LICENSE", "family": "permissive", "id": "BSD-2-Clause"}, "name": "Celery", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/redis/redis", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "in-memory store", "description": "The store most agent systems use for queues, caches and short-term state.", "kind": "memory", "id": "u-redis", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/redis/redis/HEAD/LICENSE.txt", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Redis", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/minio/minio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["file"], "inputs": ["file"]}, "what_it_is": "S3-compatible storage", "description": "Object storage you run yourself, speaking the same protocol as S3.", "kind": "tools", "id": "u-minio", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/minio/minio/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "MinIO", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/jgm/pandoc", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["PDF", "DOCX", "HTML", "Markdown"], "inputs": ["Markdown", "DOCX", "HTML"]}, "what_it_is": "document conversion", "description": "Converts between almost every document format there is. The plumbing under a lot of publishing pipelines.", "kind": "documents", "id": "u-pandoc", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Pandoc", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Kozea/WeasyPrint", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["PDF"], "inputs": ["HTML"]}, "what_it_is": "HTML to PDF", "description": "Turns HTML and CSS into a properly typeset PDF.", "kind": "documents", "id": "u-weasyprint", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Kozea/WeasyPrint/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "WeasyPrint", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/typst/typst", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["PDF"], "inputs": ["text"]}, "what_it_is": "typesetting", "description": "A typesetting system with the output quality of LaTeX and a language you can read.", "kind": "documents", "id": "u-typst", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/typst/typst/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Typst", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/playwright-mcp", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "image"], "inputs": ["URL", "text"]}, "what_it_is": "browser for agents, over MCP", "description": "Gives an agent a real browser through the Model Context Protocol. Made by Microsoft.", "kind": "tools", "id": "u-playwright-mcp", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/playwright-mcp/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Playwright MCP", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openinterpreter/openinterpreter", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code", "file"], "inputs": ["text"]}, "what_it_is": "code that runs on your machine", "description": "Lets a model write and run code locally, with the terminal as its tool.", "kind": "tools", "id": "u-openinterpreter", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openinterpreter/openinterpreter/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Open Interpreter", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/microsoft/semantic-kernel", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text", "code"]}, "what_it_is": "agent SDK", "description": "An SDK for putting models and plugins inside ordinary applications. Made by Microsoft.", "kind": "agents", "id": "u-semantickernel", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsoft/semantic-kernel/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Semantic Kernel", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openai/openai-agents-python", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "agent loop with handoffs", "description": "A small framework for agents, tools, handoffs and guardrails.", "kind": "agents", "id": "u-openaiagents", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openai/openai-agents-python/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "OpenAI Agents SDK", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Significant-Gravitas/AutoGPT", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "file"], "inputs": ["text"]}, "what_it_is": "autonomous agent platform", "description": "The project that made autonomous agents a thing, now a platform for building and running them.", "kind": "agents", "id": "u-autogpt", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Significant-Gravitas/AutoGPT/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "AutoGPT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/FoundationAgents/MetaGPT", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code", "file"], "inputs": ["text"]}, "what_it_is": "software company of agents", "description": "Assigns agents the roles of a software team and has them produce a project together.", "kind": "agents", "id": "u-metagpt", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/FoundationAgents/MetaGPT/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MetaGPT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/camel-ai/camel", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "agent society research", "description": "A framework for studying and building societies of cooperating agents.", "kind": "agents", "id": "u-camel", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/camel-ai/camel/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "CAMEL", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/letta-ai/letta", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "agents with long memory", "description": "Agents that keep memory across sessions, in a server you run yourself. Formerly MemGPT.", "kind": "agents", "id": "u-letta", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/letta-ai/letta/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Letta", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ComposioHQ/composio", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "tools for agents", "description": "Ready-made connections to hundreds of services, so an agent can act instead of only talk.", "kind": "agents", "id": "u-composio", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ComposioHQ/composio/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Composio", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/agentscope-ai/agentscope", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "multi-agent platform", "description": "A platform for building multi-agent applications, with a visual view of what each one did.", "kind": "agents", "id": "u-agentscope", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/agentscope-ai/agentscope/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "AgentScope", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Skyvern-AI/skyvern", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "image"], "inputs": ["URL", "text"]}, "what_it_is": "browser automation with vision", "description": "Automates websites using a model that looks at the page, so it survives a redesign.", "kind": "agents", "id": "u-skyvern", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/Skyvern-AI/skyvern/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Skyvern", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/browserbase/stagehand", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "text"], "inputs": ["URL", "text"]}, "what_it_is": "browser control in code", "description": "Lets you mix ordinary Playwright code with instructions written in words.", "kind": "agents", "id": "u-stagehand", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/browserbase/stagehand/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Stagehand", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Mintplex-Labs/anything-llm", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["PDF", "text"]}, "what_it_is": "desktop app over your documents", "description": "One application where you drop documents and talk to them, with agents and many model backends. Private by construction: with a local model, your documents — contracts, legal files, medical records — are read offline and never leave the machine.", "kind": "agents", "id": "u-anythingllm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Mintplex-Labs/anything-llm/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "AnythingLLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://lobehub.com", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "image"]}, "what_it_is": "workspace for a team of agents", "description": "Organises several agents as a team you hire, schedule and get reports from, self-hosted. Grew out of Lobe Chat.", "kind": "agents", "id": "u-lobechat", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "LobeHub", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/anthropics/claude-code", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "coding agent in the terminal", "description": "Anthropic's coding agent: it reads a repository, edits it and runs commands, from the terminal or an editor.", "kind": "tools", "id": "u-claudecode", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/anthropics/claude-code/HEAD/LICENSE.md", "family": "custom", "id": "custom"}, "name": "Claude Code", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it, and we still do: this shelf, this site and the tools behind it were built with a coding agent in a terminal."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/openai/codex", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "coding agent in the terminal", "description": "OpenAI's coding agent for the terminal, working on a local checkout.", "kind": "tools", "id": "u-codexcli", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/openai/codex/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Codex CLI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it. We keep it to reading rather than writing in our own work."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/google-gemini/gemini-cli", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "coding agent in the terminal", "description": "Google's open-source terminal agent for reading and changing a codebase.", "kind": "tools", "id": "u-geminicli", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/google-gemini/gemini-cli/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Gemini CLI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/aaif-goose/goose", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code", "file"], "inputs": ["code", "text"]}, "what_it_is": "local agent that uses tools", "description": "An on-machine agent built on MCP, so it works with whatever tools you give it.", "kind": "tools", "id": "u-goose", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/aaif-goose/goose/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Goose", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/RooCodeInc/Roo-Code", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "coding agent in the editor", "description": "An agent in VS Code with several modes, from asking questions to writing the change.", "kind": "tools", "id": "u-roocode", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/RooCodeInc/Roo-Code/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Roo Code", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/anomalyco/opencode", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["code", "text"]}, "what_it_is": "terminal coding agent", "description": "A terminal coding agent you point at any provider, built for keyboard use.", "kind": "tools", "id": "u-opencode", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/anomalyco/opencode/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "OpenCode", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/AntonOsika/gpt-engineer", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["text"]}, "what_it_is": "generates a codebase", "description": "Describe what you want built and it writes the project, asking about what it does not know.", "kind": "tools", "id": "u-gptengineer", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/AntonOsika/gpt-engineer/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "gpt-engineer", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11", "verdict": "We ran it."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/SWE-bench/SWE-bench", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "benchmark on real issues", "description": "Tests whether a model can fix real GitHub issues so the project's own tests pass.", "kind": "evaluation", "id": "u-swebench", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/SWE-bench/SWE-bench/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "SWE-bench", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/stackblitz-labs/bolt.diy", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["code"], "inputs": ["text"]}, "what_it_is": "build apps in the browser", "description": "Generates and runs a full web application in the browser, with any model behind it.", "kind": "tools", "id": "u-bolt", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/stackblitz-labs/bolt.diy/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "bolt.diy", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/python/mypy", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "text"], "inputs": ["code"]}, "what_it_is": "static type checker", "description": "Checks Python types before the code runs. Useful to an agent as a fast, honest verdict on what it just wrote.", "kind": "tools", "id": "u-mypy", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/python/mypy/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "mypy", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/PyCQA/bandit", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "security linter for Python", "description": "Finds common security mistakes in Python source.", "kind": "tools", "id": "u-bandit", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/PyCQA/bandit/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Bandit", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/NVIDIA/TensorRT-LLM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model", "text"]}, "what_it_is": "GPU inference engine", "description": "NVIDIA's engine for squeezing the most out of their cards when serving models.", "kind": "run locally", "id": "u-tensorrtllm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/NVIDIA/TensorRT-LLM/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "TensorRT-LLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mlc-ai/mlc-llm", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model"]}, "what_it_is": "compile models anywhere", "description": "Compiles models to run natively on phones, browsers and many kinds of GPU. That includes Android and iPhone: an offline AI assistant that runs on the phone itself, with no internet, and it does not send your data anywhere.", "kind": "run locally", "id": "u-mlcllm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mlc-ai/mlc-llm/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MLC LLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/turboderp-org/exllamav2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model", "text"]}, "what_it_is": "fast quantized inference", "description": "Runs quantized models on consumer NVIDIA cards, tuned for speed on one GPU. Local and offline — the text never leaves your machine.", "kind": "run locally", "id": "u-exllamav2", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/turboderp-org/exllamav2/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "ExLlamaV2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mozilla-ai/llamafile", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model"]}, "what_it_is": "a model as one file", "description": "Packs the model and the engine into a single executable that runs on several operating systems. One file, no installation, no internet: private AI that runs offline on a laptop.", "kind": "run locally", "id": "u-llamafile", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/mozilla-ai/llamafile/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "llamafile", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/kvcache-ai/ktransformers", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model", "text"]}, "what_it_is": "big models on small machines", "description": "Runs very large mixture-of-experts models on one GPU plus system memory. Local and offline, so even a very large model stays on your own hardware.", "kind": "run locally", "id": "u-ktransformers", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/kvcache-ai/ktransformers/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "KTransformers", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/abetlen/llama-cpp-python", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model", "text"]}, "what_it_is": "llama.cpp from Python", "description": "Python bindings for llama.cpp, with an OpenAI-compatible server included. Local and offline, and it does not send your data to any cloud service. Local and offline, with no data sent to any cloud service.", "kind": "run locally", "id": "u-llamacpppython", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/abetlen/llama-cpp-python/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "llama-cpp-python", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/xorbitsai/inference", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "vector"], "inputs": ["model", "text"]}, "what_it_is": "serve many model types", "description": "One server for language, embedding, image and audio models, with a registry of what is running. Self-hosted: the models run on your machines, not in someone else's cloud.", "kind": "run locally", "id": "u-xinference", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/xorbitsai/inference/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Xinference", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/dphnAI/sonar", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model", "text"]}, "what_it_is": "high-throughput serving", "description": "A serving engine for language and multimodal models, built for many simultaneous users on one model. Continues the project that was called Aphrodite Engine.", "kind": "run locally", "id": "u-aphrodite", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/dphnAI/sonar/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Sonar", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/intel/ipex-llm", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model", "text"]}, "what_it_is": "models on Intel hardware", "description": "Runs models on Intel GPUs and CPUs, including the integrated ones in laptops. Runs locally on ordinary Intel laptops, offline, with no cloud account.", "kind": "run locally", "id": "u-ipexllm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/intel/ipex-llm/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "IPEX-LLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/triton-inference-server/server", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["model"]}, "what_it_is": "model serving at scale", "description": "Serves models of any framework over HTTP and gRPC, with batching and versioning. Made by NVIDIA.", "kind": "run locally", "id": "u-tritonserver", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/triton-inference-server/server/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "Triton Inference Server", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ray-project/ray", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "JSON"], "inputs": ["code", "model"]}, "what_it_is": "distributed computing", "description": "Spreads training, tuning and serving across many machines from ordinary Python.", "kind": "run locally", "id": "u-ray", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ray-project/ray/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Ray", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bentoml/OpenLLM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["model"]}, "what_it_is": "run models as an API", "description": "Starts any supported open model behind an OpenAI-compatible endpoint. Made by BentoML.", "kind": "run locally", "id": "u-openllm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/bentoml/OpenLLM/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "OpenLLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/chatboxai/chatbox", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text", "image"]}, "what_it_is": "desktop chat client", "description": "A desktop client for local and remote models, on Windows, Mac and Linux.", "kind": "run locally", "id": "u-chatbox", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/chatboxai/chatbox/HEAD/LICENSE", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "Chatbox", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/text-embeddings-inference", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "embedding server", "description": "Serves embedding and reranking models fast, the way TGI serves language models. Made by Hugging Face.", "kind": "vectors", "id": "u-tei", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/text-embeddings-inference/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Text Embeddings Inference", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/michaelfeil/infinity", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text", "image"]}, "what_it_is": "embedding server", "description": "An embedding, reranking and classification server with an OpenAI-compatible API.", "kind": "vectors", "id": "u-infinity", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/michaelfeil/infinity/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Infinity", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/qdrant/fastembed", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "light embeddings", "description": "Makes embeddings without pulling in a deep learning framework. Made by Qdrant.", "kind": "vectors", "id": "u-fastembed", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/qdrant/fastembed/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "FastEmbed", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/MinishLab/model2vec", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["text"]}, "what_it_is": "tiny static embeddings", "description": "Shrinks an embedding model into a static one that is far smaller and much faster.", "kind": "vectors", "id": "u-model2vec", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/MinishLab/model2vec/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Model2Vec", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/AnswerDotAI/RAGatouille", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector", "JSON"], "inputs": ["text"]}, "what_it_is": "late interaction retrieval", "description": "Makes ColBERT-style retrieval usable without a research background.", "kind": "vectors", "id": "u-ragatouille", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/AnswerDotAI/RAGatouille/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "RAGatouille", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/supervc-stack/VectorChord", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["vector"]}, "what_it_is": "vectors inside Postgres", "description": "Vector search as a Postgres extension, so the vectors live next to the rest of your data.", "kind": "memory", "id": "u-vectorchord", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/supervc-stack/VectorChord/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "pgvecto.rs / VectorChord", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/AnswerDotAI/rerankers", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "one interface for rerankers", "description": "Wraps most reranking methods behind a single call, so you can swap them and compare.", "kind": "vectors", "id": "u-rerankers", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/AnswerDotAI/rerankers/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "rerankers", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/weaviate/Verba", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["PDF", "text"]}, "what_it_is": "RAG application", "description": "An open RAG application over Weaviate, for asking questions about your own documents.", "kind": "memory", "id": "u-verba", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/weaviate/Verba/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "Verba", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Cinnamon/kotaemon", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["PDF", "text"]}, "what_it_is": "chat with documents", "description": "A clean interface for RAG over your files, meant for end users as much as developers.", "kind": "memory", "id": "u-kotaemon", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Cinnamon/kotaemon/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "kotaemon", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/SciPhi-AI/R2R", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["PDF", "text"]}, "what_it_is": "RAG engine with an API", "description": "A retrieval engine with ingestion, graphs and permissions behind a REST API.", "kind": "memory", "id": "u-r2r", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/SciPhi-AI/R2R/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "R2R", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/JaidedAI/EasyOCR", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["image"]}, "what_it_is": "OCR in many languages", "description": "Reads text from images in over 80 languages with two lines of Python.", "kind": "documents", "id": "u-easyocr", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/JaidedAI/EasyOCR/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "EasyOCR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/RapidAI/RapidOCR", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "fast portable OCR", "description": "PaddleOCR models converted to ONNX, so they run anywhere without the original framework.", "kind": "documents", "id": "u-rapidocr", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/RapidAI/RapidOCR/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "RapidOCR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/allenai/olmocr", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "Markdown"], "inputs": ["PDF"]}, "what_it_is": "OCR for whole documents", "description": "Turns scanned documents into clean text at scale, keeping reading order. Made by Ai2.", "kind": "documents", "id": "u-olmocr", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/allenai/olmocr/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "olmOCR", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/getomni-ai/zerox", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown"], "inputs": ["PDF", "image"]}, "what_it_is": "OCR with a vision model", "description": "Feeds page images to a vision model and gets Markdown back, which handles layouts rules cannot.", "kind": "documents", "id": "u-zerox", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/getomni-ai/zerox/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Zerox", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/xavctn/img2table", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["table", "JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "tables out of images", "description": "Finds and extracts tables from scanned images and PDFs, invoices included, without a neural network. Writes them out as CSV or a dataframe.", "kind": "documents", "id": "u-img2table", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/xavctn/img2table/HEAD/LICENSE.txt", "family": "permissive", "id": "MIT"}, "name": "img2table", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pypdfium2-team/pypdfium2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image", "text"], "inputs": ["PDF"]}, "what_it_is": "PDF rendering", "description": "Renders and reads PDFs through the engine Chrome uses.", "kind": "documents", "id": "u-pdfium", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "pypdfium2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026-09-16", "why_no_figure": "Its numbers are not in our published results file, and we do not publish a figure we cannot point at.", "label": "TESTED", "machine": "Windows 11, 28 cores, CPU only", "verdict": "Ran it on our own seven-document test set. It did not make our published comparison, and we would rather say that than leave it out quietly."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/py-pdf/pypdf", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["PDF", "text"], "inputs": ["PDF"]}, "what_it_is": "PDF manipulation", "description": "Splits, merges, rotates and reads PDF files in pure Python.", "kind": "documents", "id": "u-pypdf", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/py-pdf/pypdf/HEAD/LICENSE", "family": "permissive", "id": "BSD-2-Clause"}, "name": "pypdf", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/facebookresearch/nougat", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown"], "inputs": ["PDF"]}, "what_it_is": "academic PDF to markup", "description": "Reads scientific PDFs, including the equations, into markup. Made by Meta.", "kind": "documents", "id": "u-nougat", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/facebookresearch/nougat/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Nougat", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bytedance/Dolphin", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "document image parsing", "description": "Parses a document image by first reading its layout, then each element. Made by ByteDance.", "kind": "documents", "id": "u-dolphin", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/bytedance/Dolphin/HEAD/LICENSE", "family": "custom", "id": "custom"}, "name": "Dolphin", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/facebookresearch/detectron2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "image"], "inputs": ["image"]}, "what_it_is": "detection and segmentation", "description": "Meta's library for object detection and segmentation research and production.", "kind": "vision", "id": "u-detectron2", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/facebookresearch/detectron2/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Detectron2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/open-mmlab/mmdetection", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image"]}, "what_it_is": "detection toolbox", "description": "A large collection of detection models under one training and inference interface.", "kind": "vision", "id": "u-mmdetection", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/open-mmlab/mmdetection/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MMDetection", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/roboflow/supervision", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image", "JSON"], "inputs": ["image", "video"]}, "what_it_is": "computer vision utilities", "description": "The glue around detection models: drawing, counting, tracking, filtering. Made by Roboflow.", "kind": "vision", "id": "u-supervision", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/roboflow/supervision/HEAD/LICENSE.md", "family": "permissive", "id": "MIT"}, "name": "Supervision", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mlfoundations/open_clip", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector"], "inputs": ["image", "text"]}, "what_it_is": "open CLIP models", "description": "An open implementation and set of CLIP models for matching images and text.", "kind": "vectors", "id": "u-openclip", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mlfoundations/open_clip/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "OpenCLIP", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/pytorch-image-models", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["vector", "JSON"], "inputs": ["image"]}, "what_it_is": "image model library", "description": "Hundreds of image models with pretrained weights under one interface.", "kind": "vision", "id": "u-timm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/pytorch-image-models/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "timm", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deepinsight/insightface", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "vector"], "inputs": ["image"]}, "what_it_is": "face analysis", "description": "Face detection, recognition and alignment, widely used as a building block.", "kind": "vision", "id": "u-insightface", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "InsightFace", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/serengil/deepface", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image"]}, "what_it_is": "face recognition in a line", "description": "Wraps several face recognition models behind one simple Python call.", "kind": "vision", "id": "u-deepface", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/serengil/deepface/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "DeepFace", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/kornia/kornia", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "differentiable vision", "description": "Classic computer vision operations written so they can be part of a trained network.", "kind": "vision", "id": "u-kornia", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/kornia/kornia/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Kornia", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/albumentations-team/albumentations", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "image augmentation", "description": "Fast image transformations for training, the standard choice for augmentation.", "kind": "vision", "id": "u-albumentations", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/albumentations-team/albumentations/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Albumentations", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ZhengPeng7/BiRefNet", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "high quality segmentation", "description": "Separates a subject from its background with unusually clean edges.", "kind": "vision", "id": "u-birefnet", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ZhengPeng7/BiRefNet/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "BiRefNet", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/opendatalab/DocLayout-YOLO", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["image", "PDF"]}, "what_it_is": "document layout detection", "description": "Finds the blocks of a page — text, table, figure — quickly enough for whole archives.", "kind": "documents", "id": "u-doclayoutyolo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/opendatalab/DocLayout-YOLO/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "DocLayout-YOLO", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mcmonkeyprojects/SwarmUI", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["text", "image"]}, "what_it_is": "interface over ComfyUI", "description": "A friendlier front end that still has the full node graph underneath.", "kind": "image", "id": "u-swarmui", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mcmonkeyprojects/SwarmUI/HEAD/LICENSE.txt", "family": "permissive", "id": "MIT"}, "name": "SwarmUI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Sanster/IOPaint", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "inpainting and object removal", "description": "Removes objects, erases watermarks and fills the hole, locally.", "kind": "image", "id": "u-iopaint", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Sanster/IOPaint/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "IOPaint", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/TencentARC/GFPGAN", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "face restoration", "description": "Restores faces in old or low quality photographs. Made by Tencent ARC.", "kind": "image", "id": "u-gfpgan", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/TencentARC/GFPGAN/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "GFPGAN", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/sczhou/CodeFormer", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "face restoration", "description": "Repairs badly degraded faces, with a dial between faithful and pretty.", "kind": "image", "id": "u-codeformer", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/sczhou/CodeFormer/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "CodeFormer", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/tencent-ailab/IP-Adapter", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image", "text"]}, "what_it_is": "image prompts", "description": "Lets you steer generation with a reference image instead of only words.", "kind": "image", "id": "u-ipadapter", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/tencent-ailab/IP-Adapter/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "IP-Adapter", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/instantX-research/InstantID", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image", "text"]}, "what_it_is": "keep a face across images", "description": "Generates new images of the same person from a single photograph, without training.", "kind": "image", "id": "u-instantid", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/instantX-research/InstantID/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "InstantID", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/TencentARC/PhotoMaker", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image", "text"]}, "what_it_is": "personalised portraits", "description": "Makes portraits of a given person in new styles and settings. Made by Tencent ARC.", "kind": "image", "id": "u-photomaker", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/TencentARC/PhotoMaker/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "PhotoMaker", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Acly/krita-ai-diffusion", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image", "text"]}, "what_it_is": "generation inside Krita", "description": "Puts local image generation into a real painting program, on layers.", "kind": "image", "id": "u-kritaai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/Acly/krita-ai-diffusion/HEAD/LICENSE", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "Krita AI Diffusion", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Nerogar/OneTrainer", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["image", "model"]}, "what_it_is": "train image models", "description": "One place to fine-tune and train LoRAs for image models, with a graphical interface.", "kind": "training", "id": "u-onetrainer", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/Nerogar/OneTrainer/HEAD/LICENSE.txt", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "OneTrainer", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ostris/ai-toolkit", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["image", "model"]}, "what_it_is": "training for new image models", "description": "Training scripts that keep up with the newest image and video models.", "kind": "training", "id": "u-aitoolkit", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ostris/ai-toolkit/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "AI Toolkit", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, training our own image models."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/kohya-ss/sd-scripts", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["image", "model"]}, "what_it_is": "training scripts", "description": "The training scripts most LoRA makers actually run, underneath the interfaces.", "kind": "training", "id": "u-sdscripts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/kohya-ss/sd-scripts/HEAD/LICENSE.md", "family": "permissive", "id": "Apache-2.0"}, "name": "sd-scripts", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, for training runs on image models."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/upscayl/upscayl", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "image upscaler", "description": "A desktop application that enlarges images locally, with no account.", "kind": "image", "id": "u-upscayl", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/upscayl/upscayl/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Upscayl", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/chaiNNer-org/chaiNNer", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["image"], "inputs": ["image"]}, "what_it_is": "node-based image processing", "description": "Chains image operations and upscaling models together as a graph.", "kind": "image", "id": "u-chainner", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A product that includes it must itself be open source, under the same licence.", "proof": "https://raw.githubusercontent.com/chaiNNer-org/chaiNNer/HEAD/LICENSE", "family": "strong_copyleft", "id": "GPL-3.0"}, "name": "chaiNNer", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/hpcaitech/Open-Sora", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["text", "image"]}, "what_it_is": "open video generation", "description": "An open effort to reproduce a full video generation pipeline, weights included.", "kind": "video", "id": "u-opensora", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/hpcaitech/Open-Sora/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Open-Sora", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenTalker/SadTalker", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["image", "audio"]}, "what_it_is": "talking head from a photo", "description": "Animates a single portrait to speak along with an audio track.", "kind": "video", "id": "u-sadtalker", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenTalker/SadTalker/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "SadTalker", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Rudrabha/Wav2Lip", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["video", "audio"]}, "what_it_is": "lip sync", "description": "Matches the lips in a video to a different audio track.", "kind": "video", "id": "u-wav2lip", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Wav2Lip", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/KlingAIResearch/LivePortrait", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["image", "video"]}, "what_it_is": "portrait animation", "description": "Drives a still portrait with the expressions from another video.", "kind": "video", "id": "u-liveportrait", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/KlingAIResearch/LivePortrait/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LivePortrait", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/TMElyralab/MuseTalk", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["video", "audio"]}, "what_it_is": "real-time lip sync", "description": "Lip syncing fast enough to run live. Made by Tencent Music's lab.", "kind": "video", "id": "u-musetalk", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/TMElyralab/MuseTalk/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MuseTalk", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/k4yt3x/video2x", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["video"]}, "what_it_is": "video upscaling", "description": "Enlarges and interpolates video using several models, as a desktop tool.", "kind": "video", "id": "u-video2x", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/k4yt3x/video2x/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "Video2X", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deforum/deforum-stable-diffusion", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["text"]}, "what_it_is": "animation for diffusion models", "description": "Turns image generation into animation through camera motion and prompt schedules.", "kind": "video", "id": "u-deforum", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/deforum-art/deforum-stable-diffusion/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Deforum", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/coqui-ai/TTS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "text to speech toolkit", "description": "A deep toolkit for speech synthesis and voice cloning, with many pretrained voices.", "kind": "sound", "id": "u-coquitts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can link it into a closed product, but changes to THIS library must stay open.", "proof": "https://raw.githubusercontent.com/coqui-ai/TTS/HEAD/LICENSE.txt", "family": "weak_copyleft", "id": "MPL-2.0"}, "name": "Coqui TTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/suno-ai/bark", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "generative audio", "description": "Generates speech that can also laugh, sigh and sing, plus simple sound effects.", "kind": "sound", "id": "u-bark", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/suno-ai/bark/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Bark", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/myshell-ai/OpenVoice", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text", "audio"]}, "what_it_is": "voice cloning", "description": "Copies a voice from a short sample and keeps it across languages.", "kind": "sound", "id": "u-openvoice", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/myshell-ai/OpenVoice/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "OpenVoice", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/myshell-ai/MeloTTS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "multilingual speech", "description": "Text to speech in several languages, fast enough to run on a CPU.", "kind": "sound", "id": "u-melotts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/myshell-ai/MeloTTS/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "MeloTTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/2noise/ChatTTS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "conversational speech", "description": "Speech synthesis tuned for dialogue, with pauses and fillers where a person would put them.", "kind": "sound", "id": "u-chattts", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "The strongest one here: even running it as a web service counts as distributing, so your own source has to be open too. This is the one that surprises people.", "proof": "https://raw.githubusercontent.com/2noise/ChatTTS/HEAD/LICENSE", "family": "network_copyleft", "id": "AGPL-3.0"}, "name": "ChatTTS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/yl4579/StyleTTS2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "high quality speech", "description": "A speech model that reached human-level quality on the usual benchmarks.", "kind": "sound", "id": "u-styletts2", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/yl4579/StyleTTS2/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "StyleTTS 2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/RVC-Boss/GPT-SoVITS", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text", "audio"]}, "what_it_is": "voice cloning from a minute", "description": "Clones a voice from about a minute of audio, including across languages.", "kind": "sound", "id": "u-gptsovits", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/RVC-Boss/GPT-SoVITS/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "GPT-SoVITS", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["audio"]}, "what_it_is": "voice conversion", "description": "Converts one singing or speaking voice into another.", "kind": "sound", "id": "u-rvc", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "RVC", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/NVIDIA/NeMo", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "audio"], "inputs": ["audio", "text"]}, "what_it_is": "speech and language toolkit", "description": "NVIDIA's toolkit for training and running speech recognition, synthesis and language models.", "kind": "sound", "id": "u-nemo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/NVIDIA/NeMo/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "NVIDIA NeMo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, as a speech and language toolkit."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/speechbrain/speechbrain", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "audio"], "inputs": ["audio"]}, "what_it_is": "speech toolkit", "description": "Recognition, speaker identification, enhancement and separation in one library.", "kind": "sound", "id": "u-speechbrain", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/speechbrain/speechbrain/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "SpeechBrain", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/alphacep/vosk-api", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["audio"]}, "what_it_is": "offline speech recognition", "description": "Small offline recognition models for twenty languages, including on a phone. Offline speech recognition on Android and iPhone: nothing is uploaded, and it works with no internet at all.", "kind": "sound", "id": "u-vosk", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/alphacep/vosk-api/HEAD/COPYING", "family": "permissive", "id": "Apache-2.0"}, "name": "Vosk", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deezer/spleeter", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["audio"]}, "what_it_is": "source separation", "description": "Splits a song into vocals and instruments. Made by Deezer.", "kind": "sound", "id": "u-spleeter", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/deezer/spleeter/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Spleeter", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/nomadkaraoke/python-audio-separator", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["audio"]}, "what_it_is": "stem separation", "description": "Runs the current separation models from a command line or Python.", "kind": "sound", "id": "u-audioseparator", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/nomadkaraoke/python-audio-separator/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Audio Separator", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it, splitting a track into stems."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/librosa/librosa", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "vector"], "inputs": ["audio"]}, "what_it_is": "audio analysis", "description": "The standard Python library for analysing music and audio signals.", "kind": "sound", "id": "u-librosa", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/librosa/librosa/HEAD/LICENSE.md", "family": "permissive", "id": "ISC"}, "name": "librosa", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/descriptinc/descript-audio-codec", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio", "vector"], "inputs": ["audio"]}, "what_it_is": "neural audio codec", "description": "Compresses audio to tokens a model can work with, and back again.", "kind": "sound", "id": "u-dac", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/descriptinc/descript-audio-codec/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Descript Audio Codec", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Stability-AI/stable-audio-tools", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["audio"], "inputs": ["text"]}, "what_it_is": "audio generation training", "description": "The training and inference code behind Stable Audio. Made by Stability AI.", "kind": "sound", "id": "u-stableaudiotools", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Stability-AI/stable-audio-tools/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Stable Audio Tools", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/comet-ml/opik", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "JSON"]}, "what_it_is": "LLM tracing and evaluation", "description": "Traces what an agent did and scores it, self-hosted. Made by Comet.", "kind": "evaluation", "id": "u-opik", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/comet-ml/opik/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Opik", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/truera/trulens", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "evaluate LLM apps", "description": "Instruments an application and scores each step against feedback functions.", "kind": "evaluation", "id": "u-trulens", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/truera/trulens/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "TruLens", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/evidentlyai/evidently", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "HTML"], "inputs": ["table", "text"]}, "what_it_is": "model monitoring", "description": "Watches data and model quality over time and tells you when it drifts.", "kind": "evaluation", "id": "u-evidently", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/evidentlyai/evidently/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Evidently", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Giskard-AI/giskard-oss", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON", "HTML"], "inputs": ["model", "text"]}, "what_it_is": "testing for ML and LLM", "description": "Scans a model for bias, robustness and other failures before it goes out.", "kind": "evaluation", "id": "u-giskard", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Giskard-AI/giskard/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Giskard", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/AgentOps-AI/agentops", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text", "JSON"]}, "what_it_is": "agent observability", "description": "Records agent runs, costs and failures so you can see where one went wrong.", "kind": "evaluation", "id": "u-agentops", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/AgentOps-AI/agentops/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "AgentOps", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/traceloop/openllmetry", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "OpenTelemetry for LLMs", "description": "Sends model and agent traces into the monitoring tools you already run. Made by Traceloop.", "kind": "evaluation", "id": "u-openllmetry", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/traceloop/openllmetry/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "OpenLLMetry", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/UKGovernmentBEIS/inspect_ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "evaluation framework", "description": "A framework for writing serious model evaluations, from the UK AI Safety Institute.", "kind": "evaluation", "id": "u-inspectai", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/UKGovernmentBEIS/inspect_ai/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Inspect AI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/protectai/llm-guard", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["text"]}, "what_it_is": "input and output filtering", "description": "Checks prompts and answers for injection, leaks and unsafe content. Made by Protect AI.", "kind": "evaluation", "id": "u-llmguard", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/protectai/llm-guard/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LLM Guard", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Azure/PyRIT", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "red teaming for AI", "description": "Automates attacks against a model so you find the holes before someone else does. Made by Microsoft.", "kind": "evaluation", "id": "u-pyrit", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Azure/PyRIT/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "PyRIT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/cleanlab/cleanlab", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["dataset", "table"]}, "what_it_is": "find bad labels", "description": "Finds wrong labels and low quality examples in a dataset automatically, so you fix the data instead of blaming the model.", "kind": "data", "id": "u-cleanlab", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/cleanlab/cleanlab/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Cleanlab", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Helicone/helicone", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "LLM gateway and logging", "description": "Sits in front of model providers and logs every call, with costs and caching.", "kind": "evaluation", "id": "u-helicone", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Helicone/helicone/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Helicone", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mlflow/mlflow", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "JSON"], "inputs": ["model", "table"]}, "what_it_is": "experiment tracking", "description": "Tracks runs, parameters and models, and keeps a registry of what you trained.", "kind": "data", "id": "u-mlflow", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mlflow/mlflow/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "MLflow", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/treeverse/dvc", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset", "model"], "inputs": ["dataset", "model"]}, "what_it_is": "version control for data", "description": "Puts datasets and models under version control next to the code.", "kind": "data", "id": "u-dvc", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/iterative/dvc/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "DVC", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/zenml-io/zenml", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["dataset", "model"]}, "what_it_is": "MLOps pipelines", "description": "Writes ML pipelines once and runs them on whatever infrastructure you have.", "kind": "data", "id": "u-zenml", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/zenml-io/zenml/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "ZenML", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/clearml/clearml", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "JSON"], "inputs": ["dataset", "model"]}, "what_it_is": "experiment and pipeline platform", "description": "Tracking, orchestration and data management for machine learning teams.", "kind": "data", "id": "u-clearml", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/clearml/clearml/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "ClearML", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Netflix/metaflow", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model", "JSON"], "inputs": ["dataset"]}, "what_it_is": "data science workflows", "description": "Turns a Python script into a reproducible pipeline that scales out. Made by Netflix.", "kind": "data", "id": "u-metaflow", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Netflix/metaflow/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Metaflow", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/PrefectHQ/prefect", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "workflow orchestration", "description": "Runs and retries data workflows, with a view of what failed and why.", "kind": "data", "id": "u-prefect", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/PrefectHQ/prefect/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Prefect", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/dagster-io/dagster", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["dataset"]}, "what_it_is": "data orchestration", "description": "Orchestration built around the assets you produce rather than the tasks you run.", "kind": "data", "id": "u-dagster", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/dagster-io/dagster/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Dagster", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/activeloopai/deeplake", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset"], "inputs": ["dataset"]}, "what_it_is": "dataset format for AI", "description": "Stores images, video and text in a format you can stream straight into training.", "kind": "data", "id": "u-deeplake", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/activeloopai/deeplake/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Deep Lake", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/voxel51/fiftyone", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset", "JSON"], "inputs": ["dataset", "image"]}, "what_it_is": "look at your dataset", "description": "Browse, filter and debug image and video datasets, including model mistakes.", "kind": "data", "id": "u-fiftyone", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/voxel51/fiftyone/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "FiftyOne", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/cvat-ai/cvat", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset", "JSON"], "inputs": ["image", "video"]}, "what_it_is": "annotation for vision", "description": "The widely used tool for labelling images and video, self-hosted.", "kind": "data", "id": "u-cvat", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/cvat-ai/cvat/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "CVAT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/doccano/doccano", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["dataset", "JSON"], "inputs": ["text"]}, "what_it_is": "text annotation", "description": "Labelling for text: classification, entities and sequences, in a browser.", "kind": "data", "id": "u-doccano", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/doccano/doccano/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "doccano", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/NVIDIA/Megatron-LM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["dataset"]}, "what_it_is": "large scale training", "description": "NVIDIA's code for training very large models across many GPUs.", "kind": "training", "id": "u-megatron", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/NVIDIA/Megatron-LM/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Megatron-LM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/deepspeedai/DeepSpeed", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "training optimisation", "description": "Makes large model training fit and go faster, through sharding and offloading. Made by Microsoft.", "kind": "training", "id": "u-deepspeed", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/deepspeedai/DeepSpeed/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "DeepSpeed", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/accelerate", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["code", "model"]}, "what_it_is": "run training anywhere", "description": "Runs the same training script on one GPU, many, or a cluster, without rewriting it.", "kind": "training", "id": "u-accelerate", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/accelerate/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Accelerate", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/meta-pytorch/torchtune", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "fine-tuning in PyTorch", "description": "Fine-tuning recipes written in plain PyTorch, with no framework on top.", "kind": "training", "id": "u-torchtune", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/meta-pytorch/torchtune/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "torchtune", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/verl-project/verl", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "reinforcement learning for LLMs", "description": "A library for training language models with reinforcement learning at scale.", "kind": "training", "id": "u-verl", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/volcengine/verl/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "verl", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenRLHF/OpenRLHF", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model", "dataset"]}, "what_it_is": "RLHF framework", "description": "An open framework for the reinforcement learning stage of training a model.", "kind": "training", "id": "u-openrlhf", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenRLHF/OpenRLHF/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "OpenRLHF", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bitsandbytes-foundation/bitsandbytes", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["model"]}, "what_it_is": "8 and 4 bit training", "description": "The quantization layer under most low-memory fine-tuning.", "kind": "training", "id": "u-bitsandbytes", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/bitsandbytes-foundation/bitsandbytes/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "bitsandbytes", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/huggingface/lerobot", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["video", "dataset"]}, "what_it_is": "robotics for the rest of us", "description": "Models, datasets and tools for real robots: imitation learning from demonstrations, on affordable robot arms. Made by Hugging Face.", "kind": "tools", "id": "u-lerobot", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/huggingface/lerobot/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "LeRobot", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/google-deepmind/mujoco", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video", "JSON"], "inputs": ["3D mesh", "code"]}, "what_it_is": "physics simulation", "description": "The physics engine most robot learning research runs in. Made by Google DeepMind.", "kind": "tools", "id": "u-mujoco", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/google-deepmind/mujoco/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MuJoCo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Genesis-Embodied-AI/genesis-world", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["video"], "inputs": ["3D mesh", "text"]}, "what_it_is": "generative physics engine", "description": "A very fast physics platform for robotics and embodied AI.", "kind": "tools", "id": "u-genesis", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Genesis-Embodied-AI/Genesis/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Genesis", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/Farama-Foundation/Gymnasium", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "reinforcement learning environments", "description": "The standard interface for RL environments, maintained after OpenAI Gym.", "kind": "tools", "id": "u-gymnasium", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/Farama-Foundation/Gymnasium/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Gymnasium", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/DLR-RM/stable-baselines3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["code"]}, "what_it_is": "reinforcement learning algorithms", "description": "Reliable implementations of the usual RL algorithms in PyTorch.", "kind": "tools", "id": "u-sb3", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/DLR-RM/stable-baselines3/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Stable-Baselines3", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/isaac-sim/IsaacLab", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["3D mesh"]}, "what_it_is": "robot learning on GPU", "description": "NVIDIA's framework for training robots in simulation, thousands at a time.", "kind": "tools", "id": "u-isaaclab", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/isaac-sim/IsaacLab/HEAD/LICENSE", "family": "permissive", "id": "BSD-3-Clause"}, "name": "Isaac Lab", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/bulletphysics/bullet3", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["3D mesh"]}, "what_it_is": "physics engine", "description": "A long-standing physics engine with Python bindings used widely in robotics.", "kind": "tools", "id": "u-bullet3", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "A licence of their own. Read it before you build on it; we will not summarise someone's custom terms.", "proof": "https://raw.githubusercontent.com/bulletphysics/bullet3/HEAD/LICENSE.txt", "family": "custom", "id": "custom"}, "name": "Bullet / PyBullet", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/InternLM/InternLM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "A family of open language models from the Shanghai AI Laboratory.", "kind": "models", "id": "u-internlm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/InternLM/InternLM/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "InternLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/01-ai/Yi", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "Open bilingual language models from 01.AI.", "kind": "models", "id": "u-yi", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/01-ai/Yi/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Yi", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/zai-org/GLM-4", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "open model family", "description": "An open model family from Zhipu, strong on tool use and long context.", "kind": "models", "id": "u-glm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/THUDM/GLM-4/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "GLM-4", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/MoonshotAI/Kimi-K2", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "large open model", "description": "A very large open mixture-of-experts model aimed at agent use. Made by Moonshot AI.", "kind": "models", "id": "u-kimi", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/MoonshotAI/Kimi-K2/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Kimi K2", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed here. Our own numbers are taken on hardware this does not fit, so any figure would be about the paging, not the model.", "label": "TESTED", "machine": "RTX 5070 12GB, Windows 11", "verdict": "We ran it. A large open model, and large is the operative word: it is well past what the cards we measure on can hold."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/OpenBMB/MiniCPM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "small strong models", "description": "Small models meant to run on a phone while still being useful.", "kind": "models", "id": "u-minicpm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/OpenBMB/MiniCPM/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "MiniCPM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/karpathy/nanoGPT", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["dataset"]}, "what_it_is": "smallest real training code", "description": "A few hundred lines that actually train a GPT. The clearest way to see what training is.", "kind": "models", "id": "u-nanogpt", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/karpathy/nanoGPT/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "nanoGPT", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/rasbt/LLMs-from-scratch", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["model"], "inputs": ["text"]}, "what_it_is": "build a model by hand", "description": "A book-length repository that builds a language model step by step, in plain PyTorch.", "kind": "models", "id": "u-llmscourse", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/rasbt/LLMs-from-scratch/HEAD/LICENSE.txt", "family": "permissive", "id": "Apache-2.0"}, "name": "LLMs from scratch", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/ScrapeGraphAI/Scrapegraph-ai", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["URL", "text"]}, "what_it_is": "scraping described in words", "description": "You say what you want from a page and it builds the extraction, instead of writing selectors.", "kind": "tools", "id": "u-scrapegraph", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/ScrapeGraphAI/Scrapegraph-ai/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "ScrapeGraphAI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/jina-ai/reader", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown", "text"], "inputs": ["URL"]}, "what_it_is": "any page as clean text", "description": "Turns a URL into clean text a model can read, stripping the page furniture.", "kind": "tools", "id": "u-jinareader", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/jina-ai/reader/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Jina Reader", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/matthewwithanm/python-markdownify", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["Markdown"], "inputs": ["HTML"]}, "what_it_is": "HTML to markdown", "description": "Converts HTML into Markdown, the format models read most reliably.", "kind": "tools", "id": "u-markdownify", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/matthewwithanm/python-markdownify/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Markdownify", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/AndyTheFactory/newspaper4k", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "JSON"], "inputs": ["URL", "HTML"]}, "what_it_is": "article extraction", "description": "Pulls the article, title and date out of a news page.", "kind": "tools", "id": "u-newspaper", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/AndyTheFactory/newspaper4k/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "newspaper4k", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/noamgat/lm-format-enforcer", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "force a format", "description": "Makes a model's output match a schema or grammar while it generates.", "kind": "tools", "id": "u-lmformatenforcer", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/noamgat/lm-format-enforcer/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "LM Format Enforcer", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/mangiucugna/json_repair", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["text"]}, "what_it_is": "fix broken JSON", "description": "Repairs the almost-JSON that models produce, instead of failing the whole call.", "kind": "tools", "id": "u-jsonrepair", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/mangiucugna/json_repair/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "json_repair", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/lm-sys/RouteLLM", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "route between models", "description": "Sends easy questions to a cheap model and hard ones to a strong one.", "kind": "tools", "id": "u-routellm", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/lm-sys/RouteLLM/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "RouteLLM", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/andrewyng/aisuite", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text"], "inputs": ["text"]}, "what_it_is": "one interface for providers", "description": "One call shape across model providers, so switching costs a line.", "kind": "tools", "id": "u-aisuite", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/andrewyng/aisuite/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "aisuite", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/open-webui/mcpo", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["JSON"]}, "what_it_is": "MCP servers as REST", "description": "Exposes an MCP server over plain HTTP, so tools that do not speak MCP can still use it.", "kind": "tools", "id": "u-mcpo", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/open-webui/mcpo/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "mcpo", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/modelcontextprotocol/inspector", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["JSON"]}, "what_it_is": "debug an MCP server", "description": "A visual tool for poking at an MCP server and seeing what it actually exposes.", "kind": "tools", "id": "u-mcpinspector", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "MCP Inspector", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/daytonaio/daytona", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "file"], "inputs": ["code"]}, "what_it_is": "sandboxes for agent code", "description": "Runs code an agent wrote inside an isolated environment that starts quickly.", "kind": "tools", "id": "u-daytona", "licence": {"$why": "Either the repository has no licence file where we looked, or it was not machine-readable. We would rather say so than guess.", "what_it_means": "We could not read a licence for this one. That is not the same as saying it has none - go and look before you build something you mean to sell.", "proof": null, "family": "unknown", "id": null}, "name": "Daytona", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"$what_it_means": "We ran this ourselves and formed a judgement, but we did not put a stopwatch on it — so there is no number here, on purpose. A verdict stays true on any machine; a speed taken on ours would not.", "when": "2026", "why_no_figure": "Not timed.", "label": "TESTED", "machine": "Windows 11, and our Linux server", "verdict": "We ran it, as a sandbox for code an agent writes."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/superradcompany/microsandbox", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["text", "file"], "inputs": ["code"]}, "what_it_is": "isolated code execution", "description": "Runs untrusted code in a microVM, which is a real boundary rather than a container.", "kind": "tools", "id": "u-microsandbox", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/microsandbox/microsandbox/HEAD/LICENSE", "family": "permissive", "id": "Apache-2.0"}, "name": "Microsandbox", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/fastapi/fastapi", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["code"]}, "what_it_is": "Python API framework", "description": "The framework most model and agent services are served behind, with schemas for free.", "kind": "tools", "id": "u-fastapi", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/fastapi/fastapi/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "FastAPI", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}, {"address": "https://github.com/pydantic/pydantic", "capabilities": {"$where_from": "Declared by the project that made the tool. We have not run it and have not checked it.", "outputs": ["JSON"], "inputs": ["JSON"]}, "what_it_is": "data validation", "description": "Defines and checks the shape of data. What most tool-calling code uses to trust a model's output.", "kind": "tools", "id": "u-pydantic", "licence": {"$not_legal_advice": "This is the licence file we found in the repository, not a lawyer's reading of it. What it means for YOUR product is yours to decide.", "what_it_means": "You can build a closed, commercial product on it. Keep the copyright notice.", "proof": "https://raw.githubusercontent.com/pydantic/pydantic/HEAD/LICENSE", "family": "permissive", "id": "MIT"}, "name": "Pydantic", "price": {"currency": "EUR", "value": 0.0}, "evidence": {"label": "NOT_MEASURED", "note": "We have not run this one. It is a link we collected, not something on sale."}, "state": "collected", "source": "collected_by_the_market"}]}