{"name":"LLM Releases dataset","description":"Source-backed large language model releases, lifecycle events, and model metadata.","lastModified":"2026-06-18","models":[{"id":"70ea4eef-0203-460b-8978-414b80f0cbcf","slug":"bert-large","name":"BERT","family":"BERT","version":"Large","description":"The bidirectional encoder that reshaped NLP and seeded the transformer era.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":0.34,"paramsActiveB":0.34,"isSizeDisclosed":true,"contextWindow":512,"knowledgeCutoff":"2018-03-31","status":"available","releasedDate":"2018-10-11","homepageUrl":"https://github.com/google-research/bert","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"bc1a29c0-6959-40b4-941a-05364b4b0ce1","slug":"gpt-2","name":"GPT-2","family":"GPT-2","version":"1.5B","description":"Initially withheld over misuse fears, then fully released in Nov 2019 — an early 'limited release' debate.","orgSlug":"openai","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":1.5,"paramsActiveB":1.5,"isSizeDisclosed":true,"contextWindow":1024,"knowledgeCutoff":"2019-01-31","status":"available","announcedDate":"2019-02-14","releasedDate":"2019-11-05","homepageUrl":"https://openai.com","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"c4b76945-0594-4167-b7db-c437e92b3886","slug":"gpt-3","name":"GPT-3","family":"GPT-3","version":"175B","description":"The 175B model that proved in-context learning at scale; its base API models were retired in 2024.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"dense","paramsTotalB":175,"paramsActiveB":175,"isSizeDisclosed":true,"contextWindow":2048,"knowledgeCutoff":"2019-10-31","status":"retired","releasedDate":"2020-06-11","deprecatedDate":"2023-07-06","retiredDate":"2024-01-04","homepageUrl":"https://openai.com","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"f9ce5e18-7ba3-4ccd-99ec-269a991df7b3","slug":"palm","name":"PaLM","family":"PaLM","version":"540B","description":"Google's 540B Pathways model; the API was later deprecated in favor of Gemini.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"dense","paramsTotalB":540,"paramsActiveB":540,"isSizeDisclosed":true,"knowledgeCutoff":"2021-12-31","status":"retired","releasedDate":"2022-04-04","homepageUrl":"https://ai.google","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"c9a0dcdd-3184-44e8-8be6-72de08316ce1","slug":"bloom","name":"BLOOM","family":"BLOOM","version":"176B","description":"An open, multilingual 176B model (46 languages) from a global research collaboration.","orgSlug":"bigscience","isFrontier":false,"licenseType":"open_weight","licenseName":"BigScience RAIL","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":176,"paramsActiveB":176,"isSizeDisclosed":true,"contextWindow":2048,"knowledgeCutoff":"2022-01-31","status":"available","releasedDate":"2022-07-12","homepageUrl":"https://bigscience.huggingface.co","notable":false,"org":{"id":"4992689f-9baf-429e-a966-27e82a687317","slug":"bigscience","name":"BigScience","shortName":"BigScience","orgType":"research_org","countryCode":"FR","countryName":"France"}},{"id":"a32e4ddc-764e-4aaf-b472-7ac1da4e0af4","slug":"galactica","name":"Galactica","family":"Galactica","version":"120B","description":"A science-focused model whose public demo was withdrawn after just three days over confidently wrong outputs — an early, instructive retraction.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"CC-BY-NC-4.0","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":120,"paramsActiveB":120,"isSizeDisclosed":true,"contextWindow":2048,"knowledgeCutoff":"2022-09-30","status":"withdrawn","releasedDate":"2022-11-15","retiredDate":"2022-11-17","homepageUrl":"https://galactica.org","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"2389b6f4-fc6f-4814-82ae-7db07849c51a","slug":"llama-1","name":"LLaMA","family":"Llama","version":"65B","description":"Meta's first LLaMA, released to researchers; its leak catalyzed the open-weight movement.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"LLaMA Research License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":65,"paramsActiveB":65,"isSizeDisclosed":true,"contextWindow":2048,"knowledgeCutoff":"2022-08-31","status":"available","releasedDate":"2023-02-24","homepageUrl":"https://ai.meta.com","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"c5dcbfde-41e5-49b2-8ee7-06a3fa838747","slug":"gpt-4","name":"GPT-4","family":"GPT-4","version":"4","description":"The model that brought reliable multi-step reasoning to the mainstream; size never disclosed.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":8192,"knowledgeCutoff":"2021-09-30","status":"deprecated","releasedDate":"2023-03-14","deprecatedDate":"2024-12-06","homepageUrl":"https://openai.com","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"328a942b-cf64-46b7-b938-97724fe5e3bd","slug":"claude-2","name":"Claude 2","family":"Claude 2","version":"2","description":"Anthropic's first widely-available Claude, notable for an early 100k-token context window.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":100000,"knowledgeCutoff":"2023-01-31","status":"retired","releasedDate":"2023-07-11","retiredDate":"2025-01-21","homepageUrl":"https://anthropic.com","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"2524a2e3-acd0-4cca-9e13-be2c1196aeeb","slug":"llama-2-70b","name":"Llama 2 70B","family":"Llama 2","version":"70B","description":"The release that made capable open-weight models genuinely usable for production.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 2 Community License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":70,"paramsActiveB":70,"isSizeDisclosed":true,"contextWindow":4096,"knowledgeCutoff":"2022-09-30","status":"available","releasedDate":"2023-07-18","homepageUrl":"https://ai.meta.com","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"2da94fa8-b5a0-4182-bb43-8a30277d8fef","slug":"qwen-72b","name":"Qwen-72B","family":"Qwen","version":"1","description":"Alibaba's first major open Qwen model and the start of a prolific open-weight line.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Tongyi Qianwen License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":72,"paramsActiveB":72,"isSizeDisclosed":true,"contextWindow":32000,"knowledgeCutoff":"2023-08-31","status":"available","releasedDate":"2023-11-30","homepageUrl":"https://qwen.ai","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"410689a7-238a-4067-9223-e7014363a8d9","slug":"yi-34b","name":"Yi-34B","family":"Yi","version":"34B","description":"01.AI's strong bilingual open model, with a 200k-context variant.","orgSlug":"01-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"Yi License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":34,"paramsActiveB":34,"isSizeDisclosed":true,"contextWindow":200000,"knowledgeCutoff":"2023-06-30","status":"available","releasedDate":"2023-11-06","homepageUrl":"https://www.01.ai","notable":false,"org":{"id":"69904005-2071-43ce-b0ed-c6c8e3d51246","slug":"01-ai","name":"01.AI","shortName":"01.AI","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"699398da-2ffe-41a5-940d-15f3588e61c6","slug":"falcon-180b","name":"Falcon 180B","family":"Falcon","version":"180B","description":"At launch the largest openly available model, from the UAE's TII.","orgSlug":"tii","isFrontier":false,"licenseType":"open_weight","licenseName":"Falcon-180B TII License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":180,"paramsActiveB":180,"isSizeDisclosed":true,"contextWindow":2048,"knowledgeCutoff":"2023-06-30","status":"available","releasedDate":"2023-09-06","homepageUrl":"https://falconllm.tii.ae","notable":false,"org":{"id":"47a447cd-6251-4c25-991e-2f17f8c71c8b","slug":"tii","name":"Technology Innovation Institute","shortName":"TII","orgType":"research_org","countryCode":"AE","countryName":"United Arab Emirates"}},{"id":"f97081f4-7790-4f56-afd5-57035be1d335","slug":"mistral-7b","name":"Mistral 7B","family":"Mistral 7B","version":"0.1","description":"The 7B that punched far above its weight and put Mistral on the map.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"contextWindow":8192,"knowledgeCutoff":"2023-06-30","status":"available","releasedDate":"2023-09-27","homepageUrl":"https://mistral.ai","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"85ac5545-62ca-499f-8788-7561273d1e31","slug":"mixtral-8x7b","name":"Mixtral 8x7B","family":"Mixtral","version":"8x7B","description":"The open sparse Mixture-of-Experts that brought MoE efficiency to the open ecosystem.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text"],"architecture":"moe","paramsTotalB":47,"paramsActiveB":13,"isSizeDisclosed":true,"contextWindow":32000,"knowledgeCutoff":"2023-09-30","status":"available","releasedDate":"2023-12-11","homepageUrl":"https://mistral.ai","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"7f42f529-a805-4585-bd29-b1aae2ad2111","slug":"gemini-1-0-ultra","name":"Gemini 1.0 Ultra","family":"Gemini 1","version":"1.0 Ultra","description":"Google's first natively multimodal Gemini flagship, since superseded by the 1.5/2/3 lines.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":32000,"knowledgeCutoff":"2023-04-30","status":"deprecated","releasedDate":"2023-12-06","deprecatedDate":"2025-02-15","homepageUrl":"https://deepmind.google","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"997f7a63-978c-4f41-9c03-40506dc62e92","slug":"gpt-5-6","name":"GPT-5.6","family":"GPT-5","version":"5.6","description":"OpenAI's mid-2026 flagship, headlined by an industry-leading 1.5M-token context window and long-horizon agentic tool use.","orgSlug":"openai","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio","code"],"architecture":"moe","isSizeDisclosed":false,"contextWindow":1500000,"maxOutputTokens":128000,"knowledgeCutoff":"2026-01-31","status":"preview","announcedDate":"2026-06-09","homepageUrl":"https://openai.com","notable":true,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"c6d3185c-9cef-4d76-8e32-3103c22fe8cc","slug":"llama-4-maverick","name":"Llama 4 Maverick","family":"Llama 4","version":"Maverick","description":"Meta's flagship open-weight MoE; highest MMLU among open models at release.","orgSlug":"meta","isFrontier":true,"licenseType":"open_weight","licenseName":"Llama 4 Community License","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"moe","paramsTotalB":400,"paramsActiveB":17,"isSizeDisclosed":true,"contextWindow":1000000,"maxOutputTokens":32000,"knowledgeCutoff":"2024-12-31","status":"available","releasedDate":"2025-04-05","homepageUrl":"https://ai.meta.com","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"90935faa-835b-4b77-bef3-ff6bd9977615","slug":"llama-4-scout","name":"Llama 4 Scout","family":"Llama 4","version":"Scout","description":"Efficient open-weight MoE designed for very long context on modest hardware.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 4 Community License","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"moe","paramsTotalB":109,"paramsActiveB":17,"isSizeDisclosed":true,"contextWindow":10000000,"maxOutputTokens":32000,"knowledgeCutoff":"2024-12-31","status":"available","releasedDate":"2025-04-05","homepageUrl":"https://ai.meta.com","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"5b9b2fb4-284a-4be0-9d24-b0ccb19a88e5","slug":"mistral-large-3","name":"Mistral Large 3","family":"Mistral Large","version":"3","description":"Mistral's largest open-weight MoE, aimed at frontier reasoning while remaining self-hostable.","orgSlug":"mistral","isFrontier":true,"licenseType":"open_weight","licenseName":"Mistral Research / Commercial","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"moe","paramsTotalB":675,"paramsActiveB":41,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"knowledgeCutoff":"2025-10-31","status":"available","announcedDate":"2025-12-02","releasedDate":"2025-12-02","homepageUrl":"https://mistral.ai","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"cde77355-1853-4071-9e02-036b36f21ddc","slug":"phi-4","name":"Phi-4","family":"Phi","version":"4","description":"A 14B dense model that rivals far larger ones on math and reasoning, under a permissive MIT license.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":14,"paramsActiveB":14,"isSizeDisclosed":true,"contextWindow":16000,"maxOutputTokens":16000,"knowledgeCutoff":"2024-06-30","status":"available","releasedDate":"2024-12-12","homepageUrl":"https://azure.microsoft.com/products/phi","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"d18ba23e-2152-4f79-a787-84d7dd545a74","slug":"gemma-3-27b","name":"Gemma 3 27B","family":"Gemma 3","version":"3","description":"Google's open multimodal model: 128k context, 140+ languages, runs on a single GPU.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"open_weight","licenseName":"Gemma Terms of Use","weightsAvailable":true,"modalities":["text","vision"],"architecture":"dense","paramsTotalB":27,"paramsActiveB":27,"isSizeDisclosed":true,"contextWindow":128000,"maxOutputTokens":8000,"knowledgeCutoff":"2024-08-31","status":"available","releasedDate":"2025-09-04","homepageUrl":"https://ai.google.dev/gemma","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"3f3a97ce-07d6-414b-8959-1a68f6341dcf","slug":"mistral-small-3","name":"Mistral Small 3","family":"Mistral Small","version":"3","description":"A latency-optimized 24B dense model under Apache-2.0 — a popular local-deployment workhorse.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":24,"paramsActiveB":24,"isSizeDisclosed":true,"contextWindow":32000,"maxOutputTokens":16000,"knowledgeCutoff":"2024-10-31","status":"available","releasedDate":"2025-01-30","homepageUrl":"https://mistral.ai","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"f5b0da2e-b921-4e5f-9a38-9102feb29386","slug":"falcon-3-10b","name":"Falcon 3 10B","family":"Falcon 3","version":"3","description":"UAE's TII open model designed to run on light infrastructure, including laptops.","orgSlug":"tii","isFrontier":false,"licenseType":"open_weight","licenseName":"TII Falcon License 2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":10,"paramsActiveB":10,"isSizeDisclosed":true,"contextWindow":32000,"maxOutputTokens":8000,"knowledgeCutoff":"2024-09-30","status":"available","releasedDate":"2024-12-17","homepageUrl":"https://falconllm.tii.ae","notable":false,"org":{"id":"47a447cd-6251-4c25-991e-2f17f8c71c8b","slug":"tii","name":"Technology Innovation Institute","shortName":"TII","orgType":"research_org","countryCode":"AE","countryName":"United Arab Emirates"}},{"id":"874f3d3f-6c35-4440-b13a-7974ef2380fc","slug":"falcon-h1-34b","name":"Falcon-H1 34B","family":"Falcon-H1","version":"H1","description":"A hybrid attention + state-space-model (SSM) design that matches 70B-class models with fewer parameters.","orgSlug":"tii","isFrontier":false,"licenseType":"open_weight","licenseName":"Falcon LLM License 2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":34,"paramsActiveB":34,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":16000,"knowledgeCutoff":"2025-03-31","status":"available","releasedDate":"2025-07-31","homepageUrl":"https://falcon-lm.github.io","notable":false,"org":{"id":"47a447cd-6251-4c25-991e-2f17f8c71c8b","slug":"tii","name":"Technology Innovation Institute","shortName":"TII","orgType":"research_org","countryCode":"AE","countryName":"United Arab Emirates"}},{"id":"8f320f2b-9e32-4b91-afe3-3f3c4b3f5335","slug":"jamba-1-5-large","name":"Jamba 1.5 Large","family":"Jamba","version":"1.5","description":"Israel's AI21 hybrid Mamba-Transformer MoE, with a 256k context and strong long-document throughput.","orgSlug":"ai21","isFrontier":false,"licenseType":"open_weight","licenseName":"Jamba Open Model License","weightsAvailable":true,"modalities":["text"],"architecture":"hybrid","paramsTotalB":398,"paramsActiveB":94,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":4000,"knowledgeCutoff":"2024-03-31","status":"available","releasedDate":"2024-08-22","homepageUrl":"https://www.ai21.com","notable":false,"org":{"id":"789fdb1b-121d-42db-ad8e-c92b3a82fd8f","slug":"ai21","name":"AI21 Labs","shortName":"AI21","orgType":"startup","countryCode":"IL","countryName":"Israel"}},{"id":"c176394e-f638-4261-9586-95cdf0d60862","slug":"olmo-2-32b","name":"OLMo 2 32B","family":"OLMo 2","version":"2","description":"A fully open model — weights, data, and training code all public — and the first such to beat GPT-3.5 / GPT-4o mini.","orgSlug":"allen-ai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":32,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":4000,"maxOutputTokens":4000,"knowledgeCutoff":"2024-12-31","status":"available","releasedDate":"2025-03-13","homepageUrl":"https://allenai.org","notable":false,"org":{"id":"d38a5c02-ed52-4f11-8d3c-d6b5b507086e","slug":"allen-ai","name":"Allen Institute for AI (Ai2)","shortName":"Ai2","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"6e7194da-72ee-438a-85ca-daf8fa9fa8bc","slug":"granite-3-1-8b","name":"Granite 3.1 8B","family":"Granite 3","version":"3.1","description":"IBM's enterprise-focused open model with a 128k context, Apache-2.0 licensed.","orgSlug":"ibm","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":128000,"maxOutputTokens":8000,"knowledgeCutoff":"2024-04-30","status":"available","releasedDate":"2024-12-18","homepageUrl":"https://www.ibm.com/granite","notable":false,"org":{"id":"313d3992-f81b-4751-8793-51c125444f42","slug":"ibm","name":"IBM","shortName":"IBM","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"aa08c51a-317b-4c58-9201-1ea544fd40cd","slug":"amazon-nova-pro","name":"Amazon Nova Pro","family":"Amazon Nova","version":"1.0","description":"AWS-native multimodal model with a 300k context; size and architecture undisclosed.","orgSlug":"amazon","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","video"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":300000,"maxOutputTokens":10000,"knowledgeCutoff":"2024-10-31","status":"available","releasedDate":"2024-12-03","homepageUrl":"https://aws.amazon.com/nova","notable":false,"org":{"id":"4daa4336-3d5e-42df-a61d-a6efad4c756d","slug":"amazon","name":"Amazon","shortName":"Amazon","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"56227369-9ea2-4310-935d-9fcc185c4899","slug":"command-r7b","name":"Command R7B","family":"Command","version":"R7B","description":"Cohere's smallest, fastest R-series model, tuned for RAG and tool use on modest hardware.","orgSlug":"cohere","isFrontier":false,"licenseType":"open_weight","licenseName":"CC-BY-NC 4.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":128000,"maxOutputTokens":4000,"knowledgeCutoff":"2024-04-30","status":"available","releasedDate":"2024-12-13","homepageUrl":"https://cohere.com","notable":false,"org":{"id":"17f4e087-94d7-4373-a3c2-8d365d1e0e16","slug":"cohere","name":"Cohere","shortName":"Cohere","orgType":"startup","countryCode":"CA","countryName":"Canada"}},{"id":"a288b6ea-f455-4c6b-8cf9-daf591f4c75b","slug":"gpt-4o","name":"GPT-4o","family":"GPT-4","version":"4o","description":"The 2024 omni-modal model that defined a generation of assistants. Deprecated in Feb 2026 and fully retired across ChatGPT on April 3, 2026.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":128000,"maxOutputTokens":16000,"knowledgeCutoff":"2024-06-30","status":"retired","releasedDate":"2024-05-13","deprecatedDate":"2026-02-13","retiredDate":"2026-04-03","homepageUrl":"https://openai.com","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"d8c09f21-b42a-459d-8710-767fe0abe41c","slug":"qwen3-6-27b","name":"Qwen3.6-27B","family":"Qwen3.6","version":"3.6","description":"Dense 27B that punches far above its weight on agentic coding — easy to self-host on a single GPU node.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":27,"paramsActiveB":27,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"knowledgeCutoff":"2026-01-31","status":"available","releasedDate":"2026-05-12","homepageUrl":"https://qwen.ai","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"e2a13d7e-443e-4b13-bce7-6b1252aa60b1","slug":"claude-3-7-sonnet","name":"Claude 3.7 Sonnet","family":"Claude 3","version":"3.7 Sonnet","description":"Anthropic's first hybrid-reasoning Sonnet. Shut down May 11, 2026 as the 4.x line matured.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"maxOutputTokens":64000,"knowledgeCutoff":"2024-10-31","status":"retired","releasedDate":"2025-02-24","retiredDate":"2026-05-11","homepageUrl":"https://anthropic.com","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"8fd01cfa-5dc9-461b-b434-82105fbd71b2","slug":"deepseek-v3","name":"DeepSeek-V3","family":"DeepSeek V3","version":"3","description":"The 671B/37B-active MoE release that made DeepSeek a central open-model lab before the R1 breakthrough.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":671,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2024-12-26","releasedDate":"2024-12-26","homepageUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3","paperUrl":"https://arxiv.org/abs/2412.19437","notable":true,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"45e0374d-f290-4241-baf9-1691e9165377","slug":"deepseek-r1","name":"DeepSeek-R1","family":"DeepSeek R1","version":"R1","description":"Breakout open reasoning model trained with large-scale reinforcement learning and released with weights under MIT.","orgSlug":"deepseek","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":671,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"maxOutputTokens":32768,"status":"available","announcedDate":"2025-01-20","releasedDate":"2025-01-20","homepageUrl":"https://api-docs.deepseek.com/news/news250120","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-R1","paperUrl":"https://arxiv.org/abs/2501.12948","notable":true,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"8c12af0d-a1f0-4bc4-a409-ed4266543af7","slug":"deepseek-v3-0324","name":"DeepSeek-V3-0324","family":"DeepSeek V3","version":"0324","description":"Post-R1 V3 update with improved reasoning, front-end coding, Chinese writing, search, and function calling.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":671,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-03-25","releasedDate":"2025-03-25","homepageUrl":"https://api-docs.deepseek.com/news/news250325","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3-0324","paperUrl":"https://arxiv.org/abs/2412.19437","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"78bef3f8-d4fc-4bb8-a5a9-5bef6bbf1dc8","slug":"deepseek-r1-0528","name":"DeepSeek-R1-0528","family":"DeepSeek R1","version":"0528","description":"Major R1 reasoning update with stronger math, programming, general logic, function calling, and reduced hallucinations.","orgSlug":"deepseek","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":671,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"maxOutputTokens":64000,"status":"available","announcedDate":"2025-05-28","releasedDate":"2025-05-28","homepageUrl":"https://api-docs.deepseek.com/news/news250528","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-R1-0528","paperUrl":"https://arxiv.org/abs/2501.12948","notable":true,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"742f9aef-d46f-43d9-a64e-725eb1737d17","slug":"deepseek-v3-1","name":"DeepSeek-V3.1","family":"DeepSeek V3.1","version":"3.1","description":"Hybrid thinking/non-thinking release that upgraded tool calling, long-context training, and agent task performance.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":671,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-08-21","releasedDate":"2025-08-21","homepageUrl":"https://api-docs.deepseek.com/news/news250821","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1","paperUrl":"https://arxiv.org/abs/2412.19437","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"4b0ceeaa-1f67-4cc0-bb10-d95ffbd5cd55","slug":"deepseek-v3-1-terminus","name":"DeepSeek-V3.1-Terminus","family":"DeepSeek V3.1","version":"3.1 Terminus","description":"Stability update to V3.1 focused on language consistency, code-agent reliability, and search-agent behavior.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":685,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-09-22","releasedDate":"2025-09-22","homepageUrl":"https://api-docs.deepseek.com/news/news250922","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus","paperUrl":"https://arxiv.org/abs/2412.19437","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"9a0ed29d-1dd8-44ad-a18c-f912ab0b2d3a","slug":"deepseek-v3-2-exp","name":"DeepSeek-V3.2-Exp","family":"DeepSeek V3.2","version":"3.2 Exp","description":"Experimental checkpoint that introduced DeepSeek Sparse Attention as an efficiency bridge between V3.1-Terminus and V3.2.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":685,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"preview","announcedDate":"2025-09-29","releasedDate":"2025-09-29","homepageUrl":"https://api-docs.deepseek.com/news/news250929","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp","paperUrl":"https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/DeepSeek_V3_2.pdf","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"a97f6fa5-84b4-46da-9ede-170b5930da07","slug":"deepseek-v3-2","name":"DeepSeek-V3.2","family":"DeepSeek V3.2","version":"3.2","description":"Reasoning-first agent model that adds DeepSeek Sparse Attention and thinking directly inside tool-use workflows.","orgSlug":"deepseek","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":685,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-12-01","releasedDate":"2025-12-01","homepageUrl":"https://api-docs.deepseek.com/news/news251201","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2","paperUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/resolve/main/assets/paper.pdf","notable":true,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"37f952ee-99f8-43e6-8858-9aac60fb3f9c","slug":"deepseek-v3-2-speciale","name":"DeepSeek-V3.2-Speciale","family":"DeepSeek V3.2","version":"Speciale","description":"High-compute reasoning variant of V3.2, positioned for olympiad-level math, programming, and other deep reasoning tasks.","orgSlug":"deepseek","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":685,"paramsActiveB":37,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-12-01","releasedDate":"2025-12-01","homepageUrl":"https://api-docs.deepseek.com/news/news251201","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale","paperUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3.2/resolve/main/assets/paper.pdf","notable":true,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"0f1a9831-696d-4c2a-bd81-4854b6b11395","slug":"deepseek-v4-flash","name":"DeepSeek V4-Flash","family":"DeepSeek V4","version":"4 Flash","description":"Efficient V4 companion model with 284B total / 13B active parameters and the same one-million-token context window.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":284,"paramsActiveB":13,"isSizeDisclosed":true,"contextWindow":1000000,"status":"preview","announcedDate":"2026-04-24","releasedDate":"2026-04-24","homepageUrl":"https://api-docs.deepseek.com/news/news260424","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash","paperUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash/blob/main/DeepSeek_V4.pdf","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"bcd713b2-0197-415d-89e4-4f002b087c8c","slug":"deepseek-v4-pro","name":"DeepSeek V4-Pro","family":"DeepSeek V4","version":"4 Pro","description":"Preview-series sparse MoE flagship with a one-million-token context window and 1.6T total / 49B active parameters.","orgSlug":"deepseek","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":1600,"paramsActiveB":49,"isSizeDisclosed":true,"contextWindow":1000000,"status":"preview","announcedDate":"2026-04-24","releasedDate":"2026-04-24","homepageUrl":"https://api-docs.deepseek.com/news/news260424","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro","paperUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf","notable":true,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"e1ec37f5-f87b-42d2-bbcf-a41bfaf27ea3","slug":"glm-4-5","name":"GLM-4.5","family":"GLM-4.5","version":"4.5","description":"Open agentic, reasoning, and coding foundation model that marked Z.ai international rebrand and MIT-licensed GLM push.","orgSlug":"zhipu","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":355,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-07-28","releasedDate":"2025-07-28","homepageUrl":"https://z.ai/blog/glm-4.5","modelCardUrl":"https://huggingface.co/zai-org/GLM-4.5","paperUrl":"https://arxiv.org/abs/2508.06471","notable":true,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"c3dd1056-5ac4-478a-acb2-95b66cec0e5b","slug":"glm-4-5-air","name":"GLM-4.5-Air","family":"GLM-4.5","version":"4.5 Air","description":"Compact GLM-4.5 companion with 106B total / 12B active parameters for efficient agentic reasoning and coding.","orgSlug":"zhipu","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":106,"paramsActiveB":12,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-07-28","releasedDate":"2025-07-28","homepageUrl":"https://z.ai/blog/glm-4.5","modelCardUrl":"https://huggingface.co/zai-org/GLM-4.5-Air","paperUrl":"https://arxiv.org/abs/2508.06471","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"962addde-7eb8-4fa9-afc9-886ac1358fcc","slug":"glm-4-5v","name":"GLM-4.5V","family":"GLM-V","version":"4.5V","description":"Vision-language GLM based on GLM-4.5-Air, covering image, video, document, grounding, and GUI-agent tasks.","orgSlug":"zhipu","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","vision","video"],"architecture":"moe","paramsTotalB":106,"paramsActiveB":12,"isSizeDisclosed":true,"status":"available","announcedDate":"2025-08-11","releasedDate":"2025-08-11","homepageUrl":"https://z.ai/blog/glm-4.5v","modelCardUrl":"https://huggingface.co/zai-org/GLM-4.5V","paperUrl":"https://arxiv.org/abs/2507.01006","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"2e4f6683-4d63-48e7-85d1-3dc1be2ca743","slug":"glm-4-6","name":"GLM-4.6","family":"GLM-4","version":"4.6","description":"Agentic reasoning and coding upgrade over GLM-4.5, expanding the text context window from 128K to 200K tokens.","orgSlug":"zhipu","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":357,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":200000,"status":"available","announcedDate":"2025-09-30","releasedDate":"2025-09-30","homepageUrl":"https://z.ai/blog/glm-4.6","modelCardUrl":"https://huggingface.co/zai-org/GLM-4.6","paperUrl":"https://arxiv.org/abs/2508.06471","notable":true,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"173ec83b-25ec-4c54-9b55-bab3e95dc479","slug":"glm-4-6v","name":"GLM-4.6V","family":"GLM-V","version":"4.6V","description":"Open 106B-class vision-language model with native multimodal function calling for visual agents.","orgSlug":"zhipu","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","vision","video","code"],"architecture":"moe","paramsTotalB":106,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-12-08","releasedDate":"2025-12-08","homepageUrl":"https://z.ai/blog/glm-4.6v","modelCardUrl":"https://huggingface.co/zai-org/GLM-4.6V","paperUrl":"https://arxiv.org/abs/2507.01006","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"c792d11f-4a86-48c7-8021-b178e02c3016","slug":"glm-4-7","name":"GLM-4.7","family":"GLM-4","version":"4.7","description":"Coding-focused GLM release with improved multilingual agentic coding, terminal tasks, tool use, and interface generation.","orgSlug":"zhipu","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":358,"paramsActiveB":32,"isSizeDisclosed":true,"status":"available","announcedDate":"2026-01-08","releasedDate":"2026-01-08","homepageUrl":"https://z.ai/blog/glm-4.7","modelCardUrl":"https://huggingface.co/zai-org/GLM-4.7","paperUrl":"https://arxiv.org/abs/2508.06471","notable":true,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"feb48e33-c4af-49bd-919a-40fddc53d232","slug":"glm-5","name":"GLM-5","family":"GLM-5","version":"5","description":"Z.ai flagship for complex systems engineering and long-horizon agentic tasks, scaling the GLM line to 744B total / 40B active parameters.","orgSlug":"zhipu","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":744,"paramsActiveB":40,"isSizeDisclosed":true,"status":"available","announcedDate":"2026-02-11","releasedDate":"2026-02-11","homepageUrl":"https://z.ai/blog/glm-5","modelCardUrl":"https://huggingface.co/zai-org/GLM-5","paperUrl":"https://arxiv.org/abs/2602.15763","notable":true,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"d2da1cc1-3981-433c-8a50-7a6d8a132564","slug":"glm-5-1","name":"GLM-5.1","family":"GLM-5","version":"5.1","description":"Z.ai agentic-engineering follow-up to GLM-5, with stronger coding performance and better long-horizon tool-use behavior.","orgSlug":"zhipu","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":754,"paramsActiveB":40,"isSizeDisclosed":true,"status":"available","announcedDate":"2026-04-08","releasedDate":"2026-04-08","homepageUrl":"https://z.ai/blog/glm-5.1","modelCardUrl":"https://huggingface.co/zai-org/GLM-5.1","paperUrl":"https://arxiv.org/abs/2602.15763","notable":true,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"b5d80a66-16f6-4c95-9fd2-9d4c6a5dd773","slug":"gemma-4-31b","name":"Gemma 4 31B","family":"Gemma 4","version":"31B","description":"Google DeepMind's Gemma 4 advanced-reasoning open model for personal computers, part of the April 2026 Gemma 4 family.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"dense","paramsTotalB":31,"paramsActiveB":31,"isSizeDisclosed":true,"status":"available","announcedDate":"2026-04-02","releasedDate":"2026-04-02","homepageUrl":"https://deepmind.google/models/gemma/","modelCardUrl":"https://huggingface.co/google","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"4c3f0cb1-243b-474b-bb9f-45e7d6d77567","slug":"olmo-3-think-32b","name":"OLMo 3 Think 32B","family":"OLMo 3","version":"Think 32B","description":"Ai2's fully open thinking model with public weights, code, data, checkpoints, and training details across the OLMo 3 pipeline.","orgSlug":"allen-ai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":32,"paramsActiveB":32,"isSizeDisclosed":true,"status":"available","announcedDate":"2025-12-15","releasedDate":"2025-12-15","homepageUrl":"https://allenai.org/olmo","modelCardUrl":"https://huggingface.co/allenai/Olmo-3-Think-32B","paperUrl":"https://arxiv.org/abs/2512.13961","notable":false,"org":{"id":"d38a5c02-ed52-4f11-8d3c-d6b5b507086e","slug":"allen-ai","name":"Allen Institute for AI (Ai2)","shortName":"Ai2","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"86e269c6-e90c-4c41-b96d-b2ae61ccdec9","slug":"nemotron-3-nano-30b-a3b","name":"Nemotron 3 Nano 30B-A3B","family":"Nemotron 3","version":"Nano 30B-A3B","description":"Efficient Nemotron 3 MoE checkpoint for agentic reasoning and coding, activating about 3B parameters while supporting 1M-token contexts.","orgSlug":"nvidia","isFrontier":false,"licenseType":"open_weight","licenseName":"Nemotron Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":30,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":1000000,"maxOutputTokens":16384,"status":"available","announcedDate":"2025-12-15","releasedDate":"2025-12-15","homepageUrl":"https://build.nvidia.com/nvidia/nemotron-3-nano-30b-a3b","modelCardUrl":"https://build.nvidia.com/nvidia/nemotron-3-nano-30b-a3b","paperUrl":"https://arxiv.org/abs/2512.20848","notable":false,"org":{"id":"f32419b4-52f1-4de7-86dd-1d8370296ada","slug":"nvidia","name":"NVIDIA","shortName":"NVIDIA","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"d6d8aec4-c7b1-416f-9cd3-fc0a3d74395c","slug":"nemotron-3-super-120b-a12b","name":"Nemotron 3 Super 120B-A12B","family":"Nemotron 3","version":"Super 120B-A12B","description":"Open-weight hybrid Mamba-Transformer MoE designed for collaborative agents and high-volume enterprise workflows.","orgSlug":"nvidia","isFrontier":true,"licenseType":"open_weight","licenseName":"Nemotron Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":120,"paramsActiveB":12,"isSizeDisclosed":true,"contextWindow":1000000,"maxOutputTokens":16384,"status":"available","announcedDate":"2026-03-16","releasedDate":"2026-03-16","homepageUrl":"https://build.nvidia.com/nvidia/nemotron-3-super-120b-a12b","modelCardUrl":"https://build.nvidia.com/nvidia/nemotron-3-super-120b-a12b","paperUrl":"https://arxiv.org/abs/2604.12374","notable":true,"org":{"id":"f32419b4-52f1-4de7-86dd-1d8370296ada","slug":"nvidia","name":"NVIDIA","shortName":"NVIDIA","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"fb717ddd-969d-48f8-896d-1628d897954a","slug":"nemotron-3-ultra-550b-a55b","name":"Nemotron 3 Ultra 550B-A55B","family":"Nemotron 3","version":"Ultra 550B-A55B","description":"NVIDIA's largest Nemotron 3 open-weight hybrid Mamba-Transformer MoE, tuned for agentic reasoning, coding, planning, and tool calling.","orgSlug":"nvidia","isFrontier":true,"licenseType":"open_weight","licenseName":"Nemotron Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":550,"paramsActiveB":55,"isSizeDisclosed":true,"contextWindow":1000000,"maxOutputTokens":16384,"status":"available","announcedDate":"2026-06-04","releasedDate":"2026-06-04","homepageUrl":"https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b","modelCardUrl":"https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b","paperUrl":"https://arxiv.org/abs/2512.20856","notable":true,"org":{"id":"f32419b4-52f1-4de7-86dd-1d8370296ada","slug":"nvidia","name":"NVIDIA","shortName":"NVIDIA","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"018014e8-7e4c-44f8-b4c9-f3f619320c9d","slug":"ernie-4-5-300b-a47b","name":"ERNIE-4.5-300B-A47B","family":"ERNIE 4.5","version":"300B-A47B","description":"Baidu's open ERNIE 4.5 language MoE, part of a 10-variant Apache-licensed model family built with heterogeneous multimodal MoE training.","orgSlug":"baidu","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":300,"paramsActiveB":47,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-03-16","releasedDate":"2025-06-30","homepageUrl":"https://github.com/PaddlePaddle/ERNIE","modelCardUrl":"https://huggingface.co/baidu/ERNIE-4.5-300B-A47B","notable":false,"org":{"id":"2a4d1508-7601-432d-8c75-ea2aa0da12d1","slug":"baidu","name":"Baidu","shortName":"Baidu","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"1033d952-84e0-44dc-ad86-78a5506f77dc","slug":"ernie-4-5-vl-424b-a47b","name":"ERNIE-4.5-VL-424B-A47B","family":"ERNIE 4.5","version":"VL 424B-A47B","description":"Baidu's largest ERNIE 4.5 vision-language MoE, supporting text, image, and video inputs with thinking and non-thinking modes.","orgSlug":"baidu","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","vision","video"],"architecture":"moe","paramsTotalB":424,"paramsActiveB":47,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-03-16","releasedDate":"2025-06-30","homepageUrl":"https://github.com/PaddlePaddle/ERNIE","modelCardUrl":"https://huggingface.co/baidu/ERNIE-4.5-VL-424B-A47B","notable":false,"org":{"id":"2a4d1508-7601-432d-8c75-ea2aa0da12d1","slug":"baidu","name":"Baidu","shortName":"Baidu","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"7d1d9751-5a65-4c74-b9ec-9fa4b069badc","slug":"hunyuan-large","name":"Hunyuan-Large","family":"Hunyuan","version":"Large","description":"Tencent's 389B total / 52B active open-weight Transformer MoE, released with a 256K pretraining context and 128K instruct context.","orgSlug":"tencent","isFrontier":false,"licenseType":"open_weight","licenseName":"Tencent License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":389,"paramsActiveB":52,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2024-11-04","releasedDate":"2024-11-04","homepageUrl":"https://github.com/Tencent/Tencent-Hunyuan-Large","modelCardUrl":"https://huggingface.co/tencent/Tencent-Hunyuan-Large","paperUrl":"https://arxiv.org/abs/2411.02265","notable":false,"org":{"id":"f08b4f2e-ef9e-4ef7-bf7e-be299aded92c","slug":"tencent","name":"Tencent Hunyuan","shortName":"Hunyuan","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"342f87c9-63ae-418b-b68a-fcef70aa8b7d","slug":"hunyuan-a13b-instruct","name":"Hunyuan-A13B-Instruct","family":"Hunyuan","version":"A13B","description":"Tencent Hunyuan open-weight fine-grained MoE model with 80B total parameters and 13B active parameters, optimized for agentic tool use.","orgSlug":"tencent","isFrontier":false,"licenseType":"open_weight","licenseName":"Tencent Hunyuan A13B License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":80,"paramsActiveB":13,"isSizeDisclosed":true,"status":"available","announcedDate":"2026-04-22","releasedDate":"2026-04-22","homepageUrl":"https://github.com/Tencent-Hunyuan/Hunyuan-A13B","modelCardUrl":"https://huggingface.co/tencent/Hunyuan-A13B-Instruct","notable":false,"org":{"id":"f08b4f2e-ef9e-4ef7-bf7e-be299aded92c","slug":"tencent","name":"Tencent Hunyuan","shortName":"Hunyuan","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"48560fc9-cd08-4c39-86d8-8d2ab30cffa2","slug":"seed-oss-36b-instruct","name":"Seed-OSS-36B-Instruct","family":"Seed-OSS","version":"36B Instruct","description":"ByteDance Seed's Apache-licensed long-context reasoning and agent model, with controllable thinking budgets and a native 512K context.","orgSlug":"bytedance","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":36,"paramsActiveB":36,"isSizeDisclosed":true,"contextWindow":512000,"status":"available","announcedDate":"2025-08-20","releasedDate":"2025-08-20","homepageUrl":"https://seed.bytedance.com","modelCardUrl":"https://huggingface.co/ByteDance-Seed/Seed-OSS-36B-Instruct","notable":false,"org":{"id":"5eada657-5a27-4c3f-868e-a171956035d2","slug":"bytedance","name":"ByteDance Seed","shortName":"Seed","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"f5db502b-da40-422d-8303-37a229b8f511","slug":"minimax-m1-80k","name":"MiniMax-M1-80k","family":"MiniMax M1","version":"80k","description":"Open Apache-licensed hybrid-attention reasoning model with 456B total / 45.9B active parameters and a native 1M-token context.","orgSlug":"minimax","isFrontier":true,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":456,"paramsActiveB":45.9,"isSizeDisclosed":true,"contextWindow":1000000,"maxOutputTokens":80000,"status":"available","announcedDate":"2025-06-16","releasedDate":"2025-06-16","homepageUrl":"https://www.minimax.io","modelCardUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M1-80k","paperUrl":"https://arxiv.org/abs/2506.13585","notable":true,"org":{"id":"1ba738a6-44be-4db6-afa3-59f997431b3b","slug":"minimax","name":"MiniMax","shortName":"MiniMax","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"e3c50c11-363e-43f4-aad2-6b9f100eec6c","slug":"minimax-m2-7","name":"MiniMax-M2.7","family":"MiniMax M2","version":"2.7","description":"Open-weight agentic model from MiniMax focused on real-world software engineering, office tasks, tool use, and self-improving training workflows.","orgSlug":"minimax","isFrontier":true,"licenseType":"open_weight","licenseName":"MiniMax Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":229.9,"paramsActiveB":9.8,"isSizeDisclosed":true,"status":"available","announcedDate":"2026-05-26","releasedDate":"2026-05-26","homepageUrl":"https://www.minimax.io","modelCardUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M2.7","paperUrl":"https://arxiv.org/abs/2605.26494","notable":true,"org":{"id":"1ba738a6-44be-4db6-afa3-59f997431b3b","slug":"minimax","name":"MiniMax","shortName":"MiniMax","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"c15b4980-a99d-46d6-89d4-547bcbf79c57","slug":"step-3-5-flash","name":"Step-3.5-Flash","family":"Step","version":"3.5 Flash","description":"StepFun's Apache-licensed sparse MoE model for fast agentic execution, coding, math, browsing, and tool-use workflows.","orgSlug":"stepfun","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":196,"paramsActiveB":11,"isSizeDisclosed":true,"contextWindow":256000,"status":"available","announcedDate":"2026-03-14","releasedDate":"2026-03-14","homepageUrl":"https://huggingface.co/stepfun-ai/Step-3.5-Flash","modelCardUrl":"https://huggingface.co/stepfun-ai/Step-3.5-Flash","paperUrl":"https://arxiv.org/abs/2602.10604","notable":false,"org":{"id":"aa1c8268-616e-4866-a5eb-542af5011ec4","slug":"stepfun","name":"StepFun","shortName":"StepFun","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"3e1063f0-ffe2-4b9f-aabd-9f19c36d9bbb","slug":"exaone-4-0-32b","name":"EXAONE 4.0 32B","family":"EXAONE 4","version":"4.0 32B","description":"LG AI Research's unified model with non-reasoning and reasoning modes, agentic tool use, and English, Korean, and Spanish support.","orgSlug":"lg-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"EXAONE AI Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":32,"paramsActiveB":32,"isSizeDisclosed":true,"status":"available","announcedDate":"2025-07-15","releasedDate":"2025-07-15","homepageUrl":"https://huggingface.co/LGAI-EXAONE","modelCardUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B","paperUrl":"https://arxiv.org/abs/2507.11407","notable":false,"org":{"id":"9eee4d26-483e-40c3-8351-3b54653c5294","slug":"lg-ai","name":"LG AI Research","shortName":"LG AI","orgType":"research_org","countryCode":"KR","countryName":"South Korea"}},{"id":"68d0dc47-be38-41a6-bf46-930e8d6f1e6d","slug":"gpt-5-4","name":"GPT-5.4","family":"GPT-5","version":"5.4","description":"Workhorse GPT-5 release with a dedicated Thinking mode; widely deployed across ChatGPT and the API.","orgSlug":"openai","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"moe","isSizeDisclosed":false,"contextWindow":400000,"maxOutputTokens":100000,"knowledgeCutoff":"2025-10-31","status":"available","announcedDate":"2026-03-05","releasedDate":"2026-03-05","homepageUrl":"https://openai.com","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"339701f7-c87c-476c-95b6-9477f138219b","slug":"claude-opus-4-8","name":"Claude Opus 4.8","family":"Claude Opus","version":"4.8","description":"Anthropic's most capable model, with strengthened agentic and long-running task performance.","orgSlug":"anthropic","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":500000,"maxOutputTokens":64000,"knowledgeCutoff":"2025-11-30","inputPriceUsdMtok":15,"outputPriceUsdMtok":75,"status":"available","announcedDate":"2026-05-28","releasedDate":"2026-05-28","homepageUrl":"https://anthropic.com","notable":true,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"661458eb-2b4a-4cf9-b023-5d3cffa7d6ac","slug":"claude-opus-4-6","name":"Claude Opus 4.6","family":"Claude Opus","version":"4.6","description":"Introduced genuinely autonomous multi-file coding and stronger computer use.","orgSlug":"anthropic","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"maxOutputTokens":64000,"knowledgeCutoff":"2025-08-31","inputPriceUsdMtok":15,"outputPriceUsdMtok":75,"status":"available","announcedDate":"2026-02-05","releasedDate":"2026-02-05","homepageUrl":"https://anthropic.com","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"3ee6258d-4f14-442c-acad-1a23594c9ce4","slug":"gemini-3-5-pro","name":"Gemini 3.5 Pro","family":"Gemini 3","version":"3.5 Pro","description":"Announced at Google I/O 2026; emphasizes deep multimodal reasoning over a 2M-token context.","orgSlug":"google-deepmind","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio","video","code"],"architecture":"moe","isSizeDisclosed":false,"contextWindow":2000000,"maxOutputTokens":65000,"knowledgeCutoff":"2026-01-31","status":"preview","announcedDate":"2026-05-19","homepageUrl":"https://deepmind.google","notable":true,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"4e09ac7f-8d94-4672-ace6-c130e5b3a42a","slug":"gemini-3-1-pro","name":"Gemini 3.1 Pro","family":"Gemini 3","version":"3.1 Pro","description":"Generally available multimodal flagship with native tool use and a 2M-token context.","orgSlug":"google-deepmind","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio","video","code"],"architecture":"moe","isSizeDisclosed":false,"contextWindow":2000000,"maxOutputTokens":65000,"knowledgeCutoff":"2025-09-30","status":"available","announcedDate":"2026-02-19","releasedDate":"2026-02-19","homepageUrl":"https://deepmind.google","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"9912a224-aef8-4da8-9310-5e4c5b92f275","slug":"grok-4-3","name":"Grok 4.3","family":"Grok 4","version":"4.3","description":"xAI's agentic flagship with a 1M-token context and aggressive API pricing.","orgSlug":"xai","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"moe","isSizeDisclosed":false,"contextWindow":1000000,"maxOutputTokens":64000,"knowledgeCutoff":"2025-11-30","inputPriceUsdMtok":1.25,"outputPriceUsdMtok":2.5,"status":"available","announcedDate":"2026-04-30","releasedDate":"2026-05-06","homepageUrl":"https://x.ai","notable":false,"org":{"id":"c05b8503-e7e3-4615-b271-8e928f7fd36e","slug":"xai","name":"xAI","shortName":"xAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"a50865be-a224-4695-be81-ead966833e21","slug":"qwen3-5","name":"Qwen3.5-397B","family":"Qwen3.5","version":"3.5","description":"Native vision-language MoE supporting 201 languages with a 1M-token context.","orgSlug":"alibaba","isFrontier":true,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"moe","paramsTotalB":397,"paramsActiveB":17,"isSizeDisclosed":true,"contextWindow":1000000,"maxOutputTokens":32000,"knowledgeCutoff":"2025-10-31","status":"available","announcedDate":"2026-02-20","releasedDate":"2026-02-20","homepageUrl":"https://qwen.ai","notable":true,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"b7d787b1-6a23-4a8f-a53e-fccc918bff65","slug":"mistral-medium-3-5","name":"Mistral Medium 3.5","family":"Mistral Medium","version":"3.5","description":"Dense 128B open-weight model with a 256k context and strong coding performance for its size.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_weight","licenseName":"Mistral Research / Commercial","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":128,"paramsActiveB":128,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"knowledgeCutoff":"2025-10-31","status":"available","releasedDate":"2026-03-18","homepageUrl":"https://mistral.ai","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"55f71e61-edb0-415b-a63e-c0b4a7f7e837","slug":"cohere-command-a","name":"Command A","family":"Command","version":"A","description":"Enterprise-grade model tuned for RAG, tool use, and multilingual business workloads.","orgSlug":"cohere","isFrontier":false,"licenseType":"open_weight","licenseName":"CC-BY-NC (research)","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":111,"paramsActiveB":111,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":16000,"knowledgeCutoff":"2025-06-30","status":"available","releasedDate":"2025-03-13","homepageUrl":"https://cohere.com","notable":false,"org":{"id":"17f4e087-94d7-4373-a3c2-8d365d1e0e16","slug":"cohere","name":"Cohere","shortName":"Cohere","orgType":"startup","countryCode":"CA","countryName":"Canada"}},{"id":"80004630-388d-47c7-a587-39e04df64559","slug":"yi-lightning","name":"Yi-Lightning","family":"Yi","version":"Lightning","description":"01.AI's MoE API model that reached the global top-10 on Chatbot Arena, strong in Chinese, math, and coding.","orgSlug":"01-ai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"moe","isSizeDisclosed":false,"knowledgeCutoff":"2024-07-31","status":"available","releasedDate":"2024-10-16","homepageUrl":"https://www.01.ai","notable":false,"org":{"id":"69904005-2071-43ce-b0ed-c6c8e3d51246","slug":"01-ai","name":"01.AI","shortName":"01.AI","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"cb9b540c-4d58-4214-9aa3-5f144d23de61","slug":"kimi-k2-6","name":"Kimi K2.6","family":"Kimi K2","version":"2.6","description":"Moonshot's open native multimodal agentic model for long-horizon coding, visual interface generation, and autonomous tool orchestration.","orgSlug":"moonshot","isFrontier":true,"licenseType":"open_weight","licenseName":"Modified MIT","weightsAvailable":true,"modalities":["text","vision","video","code"],"architecture":"moe","paramsTotalB":1000,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"knowledgeCutoff":"2025-10-31","status":"available","announcedDate":"2026-03-30","releasedDate":"2026-03-30","homepageUrl":"https://www.kimi.com/ai-models/kimi-k2-6","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-K2.6","paperUrl":"https://arxiv.org/abs/2602.02276","notable":true,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"b3b5029e-30a4-4065-826e-9ec447069601","slug":"jurassic-1-jumbo","name":"Jurassic-1 Jumbo","family":"Jurassic","version":"1 Jumbo","description":"AI21's first major API language model, launched through AI21 Studio.","orgSlug":"ai21","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"dense","paramsTotalB":178,"paramsActiveB":178,"isSizeDisclosed":true,"status":"retired","releasedDate":"2021-08-11","homepageUrl":"https://www.ai21.com/","notable":false,"org":{"id":"789fdb1b-121d-42db-ad8e-c92b3a82fd8f","slug":"ai21","name":"AI21 Labs","shortName":"AI21","orgType":"startup","countryCode":"IL","countryName":"Israel"}},{"id":"a7082c38-e08c-423b-b60c-8dbafb8cdbfb","slug":"yi-coder-9b","name":"Yi-Coder-9B","family":"Yi-Coder","version":"9B","description":"01.AI's compact code model trained for repository-scale programming and code completion tasks.","orgSlug":"01-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"Yi License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":9,"paramsActiveB":9,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-09-05","homepageUrl":"https://huggingface.co/01-ai/Yi-Coder-9B-Chat","modelCardUrl":"https://huggingface.co/01-ai/Yi-Coder-9B-Chat","notable":false,"org":{"id":"69904005-2071-43ce-b0ed-c6c8e3d51246","slug":"01-ai","name":"01.AI","shortName":"01.AI","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"60aaf233-4ea9-45ca-ba01-cad1aef98310","slug":"yi-1-5-34b","name":"Yi-1.5-34B","family":"Yi","version":"1.5 34B","description":"Yi 1.5 update with stronger instruction following, coding, math, and multilingual performance.","orgSlug":"01-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"Yi License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":34,"paramsActiveB":34,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2024-05-13","homepageUrl":"https://huggingface.co/01-ai/Yi-1.5-34B-Chat","modelCardUrl":"https://huggingface.co/01-ai/Yi-1.5-34B-Chat","notable":false,"org":{"id":"69904005-2071-43ce-b0ed-c6c8e3d51246","slug":"01-ai","name":"01.AI","shortName":"01.AI","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"39db2ff2-be0e-46bc-9713-26d17feda375","slug":"glm-4-9b","name":"GLM-4-9B","family":"GLM-4","version":"9B","description":"Open GLM-4 9B model family, covering chat, long-context, and code-oriented variants.","orgSlug":"zhipu","isFrontier":false,"licenseType":"open_weight","licenseName":"GLM-4 License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":9,"paramsActiveB":9,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-06-05","homepageUrl":"https://github.com/THUDM/GLM-4","modelCardUrl":"https://huggingface.co/THUDM/glm-4-9b-chat","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"4658696e-fe44-4c79-9efd-180caf4c8205","slug":"glm-4","name":"GLM-4","family":"GLM-4","version":"4","description":"Zhipu's GLM-4 flagship generation, launched as the successor to ChatGLM3 with stronger tool use and multimodal variants.","orgSlug":"zhipu","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":128000,"status":"available","releasedDate":"2024-01-16","homepageUrl":"https://open.bigmodel.cn/dev/howuse/model","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"1de89beb-d832-43e0-bf6f-0da27bed59f6","slug":"chatglm2-6b","name":"ChatGLM2-6B","family":"ChatGLM","version":"2-6B","description":"Second open ChatGLM generation, improving long context, inference efficiency, and bilingual chat quality.","orgSlug":"zhipu","isFrontier":false,"licenseType":"open_weight","licenseName":"ChatGLM2 License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":6,"paramsActiveB":6,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2023-06-25","homepageUrl":"https://github.com/THUDM/ChatGLM2-6B","modelCardUrl":"https://huggingface.co/THUDM/chatglm2-6b","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"a29b12db-5555-409e-80e1-b868acb317d1","slug":"chatglm-6b","name":"ChatGLM-6B","family":"ChatGLM","version":"6B","description":"Zhipu AI and Tsinghua KEG's first widely used open bilingual ChatGLM checkpoint.","orgSlug":"zhipu","isFrontier":false,"licenseType":"open_weight","licenseName":"ChatGLM License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":6,"paramsActiveB":6,"isSizeDisclosed":true,"contextWindow":2048,"status":"available","releasedDate":"2023-03-14","homepageUrl":"https://github.com/THUDM/ChatGLM-6B","modelCardUrl":"https://huggingface.co/THUDM/chatglm-6b","notable":false,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"2576df5a-0497-49ed-bd6c-be77bdf56c7f","slug":"llama-3-3-nemotron-super-49b","name":"Llama-3.3-Nemotron-Super-49B","family":"Nemotron","version":"Super 49B","description":"Open Llama Nemotron reasoning model from NVIDIA's 2025 Nemotron family.","orgSlug":"nvidia","isFrontier":false,"licenseType":"open_weight","licenseName":"NVIDIA Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":49,"paramsActiveB":49,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-04-02","homepageUrl":"https://huggingface.co/nvidia/Llama-3_3-Nemotron-Super-49B-v1","modelCardUrl":"https://huggingface.co/nvidia/Llama-3_3-Nemotron-Super-49B-v1","notable":false,"org":{"id":"f32419b4-52f1-4de7-86dd-1d8370296ada","slug":"nvidia","name":"NVIDIA","shortName":"NVIDIA","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"c04b743e-93af-4368-a34b-1a66291c9bf3","slug":"claude-mythos","name":"Claude Mythos","family":"Claude Mythos","version":"5","description":"A frontier model Anthropic disclosed on April 7, 2026 but declined to release publicly, citing security risk. Shipped only via 'Project Glasswing' to ~50 defensive-security partners, then suspended on June 12, 2026 under a US government directive.","orgSlug":"anthropic","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"knowledgeCutoff":"2025-11-30","status":"preview","announcedDate":"2026-04-07","homepageUrl":"https://anthropic.com","notable":true,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"ecc07e33-3103-437e-a6ff-fc44a0506aae","slug":"claude-fable-5","name":"Claude Fable 5","family":"Claude Fable","version":"5","description":"The public, guardrailed sibling of Mythos and Anthropic's most capable widely-released model, built for long-horizon agentic work. Launched June 9, 2026 across the Claude API, AWS, and Microsoft Foundry — then pulled three days later under a US government export-control directive barring access by foreign nationals.","orgSlug":"anthropic","isFrontier":true,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"knowledgeCutoff":"2025-11-30","status":"withdrawn","releasedDate":"2026-06-09","retiredDate":"2026-06-12","homepageUrl":"https://anthropic.com","notable":true,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"40824bfc-0bb8-4763-abe3-17919a7e2751","slug":"qwen3-coder-next","name":"Qwen3-Coder-Next","family":"Qwen3-Coder","version":"Next","description":"Apache-licensed Qwen3-Next coding-agent model with 80B total / 3B active parameters, 256K context, and long-horizon tool-use training.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":80,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":262144,"maxOutputTokens":65536,"status":"available","announcedDate":"2026-02-03","releasedDate":"2026-02-03","homepageUrl":"https://qwen.ai","modelCardUrl":"https://huggingface.co/Qwen/Qwen3-Coder-Next","paperUrl":"https://arxiv.org/abs/2603.00729","notable":true,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"396e7837-7930-43e4-9d32-1ed5c6cf070c","slug":"glm-5-2","name":"GLM-5.2","family":"GLM-5","version":"5.2","description":"Z.ai's latest open flagship for long-horizon coding, agentic engineering, and million-token workflows, adding IndexShare sparse-attention reuse over GLM-5.1.","orgSlug":"zhipu","isFrontier":true,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":753,"isSizeDisclosed":true,"contextWindow":1000000,"status":"available","announcedDate":"2026-06-17","releasedDate":"2026-06-17","homepageUrl":"https://z.ai/blog/glm-5.2","modelCardUrl":"https://huggingface.co/zai-org/GLM-5.2","paperUrl":"https://arxiv.org/abs/2603.12201","notable":true,"org":{"id":"901a17cd-21ae-4fdd-94cd-71bca4a3fe57","slug":"zhipu","name":"Z.ai (Zhipu AI)","shortName":"Z.ai","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"02833470-3c96-493c-a6cd-66a963494c69","slug":"minimax-m3","name":"MiniMax-M3","family":"MiniMax M3","version":"3","description":"Native multimodal MiniMax model with a one-million-token context, sparse attention, and agentic coding/cowork positioning.","orgSlug":"minimax","isFrontier":true,"licenseType":"open_weight","licenseName":"MiniMax Community License","weightsAvailable":true,"modalities":["text","vision","video","code"],"architecture":"moe","paramsTotalB":428,"paramsActiveB":23,"isSizeDisclosed":true,"contextWindow":1000000,"status":"available","announcedDate":"2026-06-16","releasedDate":"2026-06-16","homepageUrl":"https://www.minimax.io","modelCardUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M3","paperUrl":"https://arxiv.org/abs/2606.13392","notable":true,"org":{"id":"1ba738a6-44be-4db6-afa3-59f997431b3b","slug":"minimax","name":"MiniMax","shortName":"MiniMax","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"63e2545b-af2b-4c00-ae72-98b745b32c01","slug":"sarvam-105b","name":"Sarvam-105B","family":"Sarvam","version":"105B","description":"Apache-licensed Indian-context MoE from Sarvam AI, optimized for reasoning, coding, agentic tasks, and 22 Indian languages.","orgSlug":"sarvam","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":105,"paramsActiveB":10.3,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2026-03-06","releasedDate":"2026-03-06","homepageUrl":"https://www.sarvam.ai/blogs/sarvam-30b-105b","modelCardUrl":"https://huggingface.co/sarvamai/sarvam-105b","notable":true,"org":{"id":"271c4a05-6c35-466d-8b07-9531196e96ab","slug":"sarvam","name":"Sarvam AI","shortName":"Sarvam","orgType":"startup","countryCode":"IN","countryName":"India"}},{"id":"5c1f458e-71d2-450a-b0ed-0ac1ce004c02","slug":"kimi-k1-5","name":"Kimi k1.5","family":"Kimi","version":"k1.5","description":"Moonshot's multimodal reinforcement-learning reasoning model, reported as matching OpenAI o1 on math, coding, and multimodal reasoning.","orgSlug":"moonshot","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","announcedDate":"2025-01-20","releasedDate":"2025-01-20","homepageUrl":"https://arxiv.org/abs/2501.12599","paperUrl":"https://arxiv.org/abs/2501.12599","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"901fe5dc-1ae1-4fcf-84ba-b67076ee04ba","slug":"moonlight-16b-a3b-instruct","name":"Moonlight-16B-A3B-Instruct","family":"Moonlight","version":"16B-A3B Instruct","description":"MIT-licensed 16B/3B-active MoE trained with Moonshot's scalable Muon optimizer experiments.","orgSlug":"moonshot","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":16,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":8000,"status":"available","announcedDate":"2025-02-24","releasedDate":"2025-02-24","homepageUrl":"https://huggingface.co/moonshotai/Moonlight-16B-A3B-Instruct","modelCardUrl":"https://huggingface.co/moonshotai/Moonlight-16B-A3B-Instruct","paperUrl":"https://arxiv.org/abs/2502.16982","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"12caf265-3448-4806-bbac-61c0cf41b9bf","slug":"kimi-vl-a3b-instruct","name":"Kimi-VL-A3B-Instruct","family":"Kimi-VL","version":"A3B Instruct","description":"Efficient MIT-licensed vision-language MoE for OCR, image/video understanding, long documents, and OS-style agent tasks.","orgSlug":"moonshot","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","vision","video"],"architecture":"moe","paramsTotalB":16,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-04-17","releasedDate":"2025-04-17","homepageUrl":"https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct","paperUrl":"https://arxiv.org/abs/2504.07491","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"e31432ae-ac21-412b-bba0-eedc6352f1df","slug":"kimi-audio-7b-instruct","name":"Kimi-Audio-7B-Instruct","family":"Kimi-Audio","version":"7B Instruct","description":"Open audio foundation model for audio understanding, generation, speech recognition, audio QA, captioning, and speech conversation.","orgSlug":"moonshot","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","audio"],"architecture":"hybrid","paramsTotalB":10,"paramsActiveB":10,"isSizeDisclosed":true,"status":"available","announcedDate":"2025-04-25","releasedDate":"2025-04-25","homepageUrl":"https://huggingface.co/moonshotai/Kimi-Audio-7B-Instruct","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-Audio-7B-Instruct","paperUrl":"https://arxiv.org/abs/2504.18425","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"eb5d01af-1a0f-4eb0-9fb6-ae7fe27eec48","slug":"kimi-dev-72b","name":"Kimi-Dev-72B","family":"Kimi-Dev","version":"72B","description":"MIT-licensed coding LLM trained with repository-level reinforcement learning for software issue resolution.","orgSlug":"moonshot","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":73,"paramsActiveB":73,"isSizeDisclosed":true,"status":"available","announcedDate":"2025-06-17","releasedDate":"2025-06-17","homepageUrl":"https://huggingface.co/moonshotai/Kimi-Dev-72B","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-Dev-72B","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"3ee6349c-a5cf-40ff-aa84-0ed34d1bb8dd","slug":"kimi-vl-a3b-thinking-2506","name":"Kimi-VL-A3B-Thinking-2506","family":"Kimi-VL","version":"A3B Thinking 2506","description":"Updated MIT-licensed Kimi-VL reasoning model with better multimodal reasoning, video understanding, high-resolution perception, and lower thinking-token use.","orgSlug":"moonshot","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","vision","video"],"architecture":"moe","paramsTotalB":16,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","announcedDate":"2025-06-21","releasedDate":"2025-06-21","homepageUrl":"https://huggingface.co/moonshotai/Kimi-VL-A3B-Thinking-2506","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-VL-A3B-Thinking-2506","paperUrl":"https://arxiv.org/abs/2504.07491","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"74e6143e-6730-474c-b07b-f6f6cdef09ab","slug":"kimi-k2-instruct","name":"Kimi K2 Instruct","family":"Kimi K2","version":"Instruct","description":"Original open K2 post-trained model: a 1T-parameter MoE optimized for coding, reasoning, and tool-using agentic workflows.","orgSlug":"moonshot","isFrontier":true,"licenseType":"open_weight","licenseName":"Modified MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":1000,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":128000,"maxOutputTokens":32000,"status":"available","announcedDate":"2025-07-11","releasedDate":"2025-07-11","homepageUrl":"https://www.kimi.com/blog/kimi-k2","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","paperUrl":"https://arxiv.org/abs/2507.20534","notable":true,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"094ea69e-f16e-444e-bc6b-ee699615d061","slug":"kimi-k2-instruct-0905","name":"Kimi K2 Instruct 0905","family":"Kimi K2","version":"Instruct 0905","description":"September 2025 K2 update with stronger agentic coding, better frontend generation, and a doubled 256K context window.","orgSlug":"moonshot","isFrontier":true,"licenseType":"open_weight","licenseName":"Modified MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":1000,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"status":"available","announcedDate":"2025-09-05","releasedDate":"2025-09-05","homepageUrl":"https://moonshotai.github.io/Kimi-K2/","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct-0905","paperUrl":"https://arxiv.org/abs/2507.20534","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"09fb08a7-a439-4e22-b1f4-4bc070d4ed02","slug":"kimi-linear-48b-a3b-instruct","name":"Kimi-Linear-48B-A3B-Instruct","family":"Kimi Linear","version":"48B-A3B Instruct","description":"MIT-licensed hybrid linear-attention model using Kimi Delta Attention, built for million-token contexts with much lower KV-cache usage.","orgSlug":"moonshot","isFrontier":false,"licenseType":"open_source","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":48,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":1048576,"status":"available","announcedDate":"2025-10-31","releasedDate":"2025-10-31","homepageUrl":"https://huggingface.co/moonshotai/Kimi-Linear-48B-A3B-Instruct","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-Linear-48B-A3B-Instruct","paperUrl":"https://arxiv.org/abs/2510.26692","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"bb9d1b25-6492-4a62-9372-270797fd0cf4","slug":"kimi-k2-thinking","name":"Kimi K2 Thinking","family":"Kimi K2","version":"Thinking","description":"Open K2 reasoning-agent variant that interleaves step-by-step thinking with tool calls and supports stable 200-300 step tool-use trajectories.","orgSlug":"moonshot","isFrontier":true,"licenseType":"open_weight","licenseName":"Modified MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":1000,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"status":"available","announcedDate":"2025-11-06","releasedDate":"2025-11-06","homepageUrl":"https://www.kimi.com/blog/kimi-k2-thinking","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-K2-Thinking","notable":true,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"f71a5812-6348-4010-acb7-dcf177b05111","slug":"kimi-k2-5","name":"Kimi K2.5","family":"Kimi K2","version":"2.5","description":"Open multimodal Kimi model that adds native visual agentic intelligence, instant and thinking modes, and agent-swarm workflows on top of the K2 base.","orgSlug":"moonshot","isFrontier":true,"licenseType":"open_weight","licenseName":"Modified MIT","weightsAvailable":true,"modalities":["text","vision","video","code"],"architecture":"moe","paramsTotalB":1000,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":256000,"maxOutputTokens":32000,"status":"available","announcedDate":"2026-01-27","releasedDate":"2026-01-27","homepageUrl":"https://www.kimi.com/ai-models/kimi-k2-5","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-K2.5","paperUrl":"https://arxiv.org/abs/2602.02276","notable":true,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"afb807db-ac54-493c-b4bf-7e26d37fd7d4","slug":"kimi-k2-7-code","name":"Kimi K2.7 Code","family":"Kimi K2","version":"2.7 Code","description":"Moonshot's open coding-focused agentic model built on K2.6, with native vision/video input, forced thinking mode, and stronger long-horizon software-engineering performance.","orgSlug":"moonshot","isFrontier":true,"licenseType":"open_weight","licenseName":"Modified MIT","weightsAvailable":true,"modalities":["text","vision","video","code"],"architecture":"moe","paramsTotalB":1000,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":262144,"maxOutputTokens":32000,"inputPriceUsdMtok":0.95,"outputPriceUsdMtok":4,"status":"available","announcedDate":"2026-06-18","releasedDate":"2026-06-18","homepageUrl":"https://www.kimi.com/resources/kimi-k2-7-code","modelCardUrl":"https://huggingface.co/moonshotai/Kimi-K2.7-Code","notable":true,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"75582deb-3f6c-467f-81cf-76ba02ef1ce5","slug":"o3","name":"OpenAI o3","family":"o3","version":"3","description":"Reasoning model released alongside o4-mini with tool use, image reasoning, and stronger agentic problem solving.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2025-04-16","homepageUrl":"https://openai.com/index/introducing-o3-and-o4-mini/","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"a8a0eb97-8274-4661-afc4-3ec6ecfb218d","slug":"gpt-4-1","name":"GPT-4.1","family":"GPT-4.1","version":"4.1","description":"API model family focused on coding, instruction following, and one-million-token long-context work.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":1000000,"maxOutputTokens":32768,"knowledgeCutoff":"2024-06-30","status":"deprecated","releasedDate":"2025-04-14","homepageUrl":"https://openai.com/index/gpt-4-1/","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"071fbfe6-e4d6-487e-852e-269e43d6cfca","slug":"o1","name":"OpenAI o1","family":"o1","version":"1","description":"General release of OpenAI's o1 reasoning model with stronger deliberative reasoning and multimodal ChatGPT integration.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"deprecated","releasedDate":"2024-12-05","homepageUrl":"https://openai.com/index/introducing-openai-o1/","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"9b5600c5-dab0-4ebb-8a72-c5c473ea14ff","slug":"o1-preview","name":"OpenAI o1-preview","family":"o1","version":"preview","description":"OpenAI's first public reasoning-model preview, optimized to spend more inference time on hard math, coding, and science tasks.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"retired","releasedDate":"2024-09-12","homepageUrl":"https://openai.com/index/introducing-openai-o1-preview/","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"d60ce9af-0d8c-4afe-8647-e27b55c63451","slug":"gpt-4-turbo","name":"GPT-4 Turbo","family":"GPT-4","version":"Turbo","description":"Lower-cost GPT-4 generation with a 128K context window, introduced at OpenAI DevDay.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":128000,"knowledgeCutoff":"2023-04-30","status":"deprecated","releasedDate":"2023-11-06","homepageUrl":"https://openai.com/index/new-models-and-developer-products-announced-at-devday/","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"bfa0a008-7ec4-4c6f-8ca3-583d10be40cf","slug":"gpt-3-5-turbo","name":"GPT-3.5 Turbo","family":"GPT-3.5","version":"Turbo","description":"OpenAI's first ChatGPT API model, bringing the GPT-3.5 chat-tuned line to developers at much lower cost than text-davinci-003.","orgSlug":"openai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":4096,"status":"retired","releasedDate":"2023-03-01","homepageUrl":"https://openai.com/index/introducing-chatgpt-and-whisper-apis/","notable":false,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"53fbba14-d8c0-48ab-bc2d-e17672791c93","slug":"claude-opus-4","name":"Claude Opus 4","family":"Claude 4","version":"Opus 4","description":"First Claude 4 Opus model, positioned for long-running agentic and coding work before the 4.x point releases.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"status":"deprecated","releasedDate":"2025-05-22","homepageUrl":"https://www.anthropic.com/news/claude-4","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"eec6c0ef-84c9-4888-9261-ec57a3a8175f","slug":"claude-3-5-haiku","name":"Claude 3.5 Haiku","family":"Claude 3.5","version":"Haiku","description":"Fast, lower-cost Claude 3.5 model for latency-sensitive coding, tool-use, and customer-facing workloads.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"status":"deprecated","releasedDate":"2024-10-22","homepageUrl":"https://www.anthropic.com/news/3-5-models-and-computer-use","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"09afa0ae-77b8-4678-9920-aa7ea51146b5","slug":"claude-3-5-sonnet-20240620","name":"Claude 3.5 Sonnet","family":"Claude 3.5","version":"Sonnet 20240620","description":"Major Sonnet upgrade that became Anthropic's default high-intelligence workhorse for coding, writing, and visual reasoning.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"status":"retired","releasedDate":"2024-06-20","homepageUrl":"https://www.anthropic.com/news/claude-3-5-sonnet","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"6e20e94a-64c6-4f93-b766-4d3f254ab993","slug":"claude-3-opus","name":"Claude 3 Opus","family":"Claude 3","version":"Opus","description":"Highest-capability Claude 3 model, launched with Sonnet and Haiku and Anthropic's first major vision-capable Claude family.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"status":"deprecated","releasedDate":"2024-03-04","homepageUrl":"https://www.anthropic.com/news/claude-3-family","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"a2d4d913-8738-4cd2-a7c2-838d3e88020a","slug":"claude-2-1","name":"Claude 2.1","family":"Claude 2","version":"2.1","description":"Claude update with a 200K context window, lower hallucination rates, and improved tool-use beta support.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":200000,"status":"retired","releasedDate":"2023-11-21","homepageUrl":"https://www.anthropic.com/news/claude-2-1","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"b01c725d-5f13-4bb1-a6f3-1bc47663c5c9","slug":"claude-1","name":"Claude 1","family":"Claude","version":"1","description":"Anthropic's first broadly announced Claude assistant model, launched through an API and select product partners.","orgSlug":"anthropic","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"status":"retired","releasedDate":"2023-03-14","homepageUrl":"https://www.anthropic.com/news/introducing-claude","notable":false,"org":{"id":"ccc838dc-a6d5-4891-9706-c38ce0bd1d4f","slug":"anthropic","name":"Anthropic","shortName":"Anthropic","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"45f69261-5a2e-40f8-8db4-8456381d1d5c","slug":"gemini-2-5-pro","name":"Gemini 2.5 Pro","family":"Gemini 2.5","version":"Pro","description":"Reasoning-focused Gemini 2.5 model that made thinking a core part of Google's flagship model line.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio","video","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":1000000,"status":"deprecated","releasedDate":"2025-03-25","homepageUrl":"https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"c5b0602d-362e-41a8-9d4c-361b5213f299","slug":"gemini-2-0-flash","name":"Gemini 2.0 Flash","family":"Gemini 2","version":"2.0 Flash","description":"First Gemini 2.0 release, built for native multimodal input/output, tool use, and agentic product integrations.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio","video","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":1000000,"status":"deprecated","releasedDate":"2024-12-11","homepageUrl":"https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"5787224b-116f-4d76-90b6-695ba4f3d97e","slug":"gemma-2-27b","name":"Gemma 2 27B","family":"Gemma 2","version":"27B","description":"Second-generation Gemma model, improving open-weight quality and efficiency at 9B and 27B sizes.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"open_weight","licenseName":"Gemma Terms of Use","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":27,"paramsActiveB":27,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2024-06-27","homepageUrl":"https://blog.google/technology/developers/google-gemma-2/","modelCardUrl":"https://huggingface.co/google/gemma-2-27b","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"11893725-e33c-4b34-ab8e-0ceafd1b0071","slug":"gemma-7b","name":"Gemma 7B","family":"Gemma","version":"7B","description":"First Gemma open-weight text model family, derived from the same research lineage as Gemini.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"open_weight","licenseName":"Gemma Terms of Use","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2024-02-21","homepageUrl":"https://blog.google/technology/developers/gemma-open-models/","modelCardUrl":"https://huggingface.co/google/gemma-7b","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"65cf62c2-fd19-46f6-89fd-f70f8d462f83","slug":"gemini-1-5-pro","name":"Gemini 1.5 Pro","family":"Gemini 1.5","version":"Pro","description":"Gemini generation that introduced production-scale long context, eventually expanding to a two-million-token window.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","audio","video","code"],"architecture":"moe","isSizeDisclosed":false,"contextWindow":2000000,"status":"deprecated","releasedDate":"2024-02-15","homepageUrl":"https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"f77e5ed4-7d56-4bf8-a06e-f2b54b575e8a","slug":"palm-2","name":"PaLM 2","family":"PaLM","version":"2","description":"Google's improved multilingual, reasoning, and coding foundation model family introduced at I/O 2023.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"dense","isSizeDisclosed":false,"status":"retired","releasedDate":"2023-05-10","homepageUrl":"https://blog.google/technology/ai/google-palm-2-ai-large-language-model/","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"da7321b3-c201-4e05-a027-389a1a8dace1","slug":"llama-3-3-70b","name":"Llama 3.3 70B","family":"Llama 3.3","version":"70B","description":"Late-2024 70B Llama update delivering much of the 405B instruction-following quality at lower serving cost.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.3 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":70,"paramsActiveB":70,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-12-06","homepageUrl":"https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct","modelCardUrl":"https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"8d486b43-af28-464f-8631-d5b3017e8f4f","slug":"llama-3-2-90b-vision","name":"Llama 3.2 90B Vision","family":"Llama 3.2","version":"90B Vision","description":"First Llama family release with native vision models, alongside smaller edge-oriented 1B and 3B text models.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.2 Community License","weightsAvailable":true,"modalities":["text","vision"],"architecture":"dense","paramsTotalB":90,"paramsActiveB":90,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-09-25","homepageUrl":"https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/","modelCardUrl":"https://huggingface.co/meta-llama/Llama-3.2-90B-Vision-Instruct","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"869a64a7-f6d6-4f5b-ae89-5f5bf9792dc0","slug":"llama-3-1-405b","name":"Llama 3.1 405B","family":"Llama 3.1","version":"405B","description":"Meta's first frontier-scale open Llama model, with 405B parameters, 128K context, multilingual support, and tool-use improvements.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.1 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":405,"paramsActiveB":405,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-07-23","homepageUrl":"https://ai.meta.com/blog/meta-llama-3-1/","modelCardUrl":"https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct","notable":true,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"5d6bd8be-2ddc-423a-897c-7e7a7cdbe0d3","slug":"llama-3-70b","name":"Llama 3 70B","family":"Llama 3","version":"70B","description":"First Llama 3 release, with 8B and 70B open models and a stronger tokenizer, data mix, and post-training stack.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":70,"paramsActiveB":70,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2024-04-18","homepageUrl":"https://ai.meta.com/blog/meta-llama-3/","modelCardUrl":"https://huggingface.co/meta-llama/Meta-Llama-3-70B-Instruct","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"97669bb9-88af-4113-a8e4-414cd998146f","slug":"code-llama-34b","name":"Code Llama 34B","family":"Code Llama","version":"34B","description":"Meta's first code-specialized Llama model family, released in base, Python, and instruction-tuned variants.","orgSlug":"meta","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 2 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":34,"paramsActiveB":34,"isSizeDisclosed":true,"contextWindow":16000,"status":"available","releasedDate":"2023-08-24","homepageUrl":"https://ai.meta.com/blog/code-llama-large-language-model-coding/","modelCardUrl":"https://huggingface.co/codellama/CodeLlama-34b-Instruct-hf","notable":false,"org":{"id":"5c9b3937-91d3-4e9f-8d99-b09f95c9442c","slug":"meta","name":"Meta AI","shortName":"Meta","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"e2027255-9665-45ae-a3cb-6c2f4b618084","slug":"grok-4","name":"Grok 4","family":"Grok","version":"4","description":"xAI's fourth-generation Grok line, preceding the later 4.x API updates already tracked in the catalog.","orgSlug":"xai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"deprecated","releasedDate":"2025-07-09","homepageUrl":"https://x.ai/news/grok-4","notable":false,"org":{"id":"c05b8503-e7e3-4615-b271-8e928f7fd36e","slug":"xai","name":"xAI","shortName":"xAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"9408d080-fadb-4652-888b-8d09b3d496e2","slug":"grok-3","name":"Grok 3","family":"Grok","version":"3","description":"xAI's third-generation model family, introduced with stronger reasoning, search, and coding modes.","orgSlug":"xai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"deprecated","releasedDate":"2025-02-17","homepageUrl":"https://x.ai/news/grok-3","notable":false,"org":{"id":"c05b8503-e7e3-4615-b271-8e928f7fd36e","slug":"xai","name":"xAI","shortName":"xAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"a8d92eb0-8e16-4a46-b5a9-da7896dac997","slug":"grok-2","name":"Grok-2","family":"Grok","version":"2","description":"Second-generation Grok release with Grok-2 and Grok-2 mini for chat, coding, reasoning, and image-enabled product experiences.","orgSlug":"xai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"retired","releasedDate":"2024-08-13","homepageUrl":"https://x.ai/news/grok-2","notable":false,"org":{"id":"c05b8503-e7e3-4615-b271-8e928f7fd36e","slug":"xai","name":"xAI","shortName":"xAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"9976cddb-878d-4f40-8c34-e15c987c13b8","slug":"grok-1-5","name":"Grok-1.5","family":"Grok","version":"1.5","description":"Grok update with stronger reasoning and a 128K context window.","orgSlug":"xai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":128000,"status":"retired","releasedDate":"2024-03-28","homepageUrl":"https://x.ai/news/grok-1.5","notable":false,"org":{"id":"c05b8503-e7e3-4615-b271-8e928f7fd36e","slug":"xai","name":"xAI","shortName":"xAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"4e8394ac-a7da-4740-a00d-20dad72a0021","slug":"grok-1","name":"Grok-1","family":"Grok","version":"1","description":"xAI's first Grok model, later released as open weights with a 314B-parameter MoE checkpoint.","orgSlug":"xai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":314,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-11-04","homepageUrl":"https://x.ai/news/grok","modelCardUrl":"https://github.com/xai-org/grok-1","notable":false,"org":{"id":"c05b8503-e7e3-4615-b271-8e928f7fd36e","slug":"xai","name":"xAI","shortName":"xAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"77ab02e4-82ba-45f2-8fd5-16fd308cc82b","slug":"magistral-medium","name":"Magistral Medium","family":"Magistral","version":"Medium","description":"Mistral's first dedicated reasoning model family, released in Small open-weight and Medium enterprise/API tiers.","orgSlug":"mistral","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2025-06-10","homepageUrl":"https://mistral.ai/news/magistral/","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"604bedf7-7d99-492e-88a5-40eeea82da06","slug":"mistral-small-3-1","name":"Mistral Small 3.1","family":"Mistral Small","version":"3.1","description":"Apache-licensed Small update adding vision and a 128K context window to the efficient 24B line.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"dense","paramsTotalB":24,"paramsActiveB":24,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-03-17","homepageUrl":"https://mistral.ai/news/mistral-small-3-1/","modelCardUrl":"https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"fe82579d-a979-4fca-8b74-c5985e4a3522","slug":"ministral-8b","name":"Ministral 8B","family":"Ministral","version":"8B","description":"Small Mistral model line optimized for edge and low-latency workloads.","orgSlug":"mistral","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-10-16","homepageUrl":"https://mistral.ai/news/ministraux/","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"1c4f6c88-c335-40dc-b740-3f04330aa248","slug":"pixtral-12b","name":"Pixtral 12B","family":"Pixtral","version":"12B","description":"Mistral's first open multimodal model, adding image understanding to a Mistral text backbone.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","vision"],"architecture":"dense","paramsTotalB":12,"paramsActiveB":12,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-09-17","homepageUrl":"https://mistral.ai/news/pixtral-12b/","modelCardUrl":"https://huggingface.co/mistralai/Pixtral-12B-2409","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"7140757f-018b-4b78-aa10-c15054fa44a0","slug":"mistral-nemo","name":"Mistral NeMo","family":"Mistral NeMo","version":"12B","description":"Apache-licensed 12B model co-developed with NVIDIA, including a 128K context window and strong multilingual tokenization.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":12,"paramsActiveB":12,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-07-18","homepageUrl":"https://mistral.ai/news/mistral-nemo/","modelCardUrl":"https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"1aab1741-2b58-4a7b-99c2-04e98d6de054","slug":"codestral-22b","name":"Codestral 22B","family":"Codestral","version":"22B","description":"Mistral's first code-specialized model, trained for code generation, fill-in-the-middle, and multi-language programming tasks.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_weight","licenseName":"Mistral AI Non-Production License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":22,"paramsActiveB":22,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2024-05-29","homepageUrl":"https://mistral.ai/news/codestral/","modelCardUrl":"https://huggingface.co/mistralai/Codestral-22B-v0.1","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"0b1397b2-db85-4ab2-890f-4712e9406c7e","slug":"mixtral-8x22b","name":"Mixtral 8x22B","family":"Mixtral","version":"8x22B","description":"Larger open Mixtral sparse MoE with 141B total and 39B active parameters, released under Apache-2.0.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":141,"paramsActiveB":39,"isSizeDisclosed":true,"contextWindow":64000,"status":"available","releasedDate":"2024-04-17","homepageUrl":"https://mistral.ai/news/mixtral-8x22b/","modelCardUrl":"https://huggingface.co/mistralai/Mixtral-8x22B-Instruct-v0.1","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"a414422d-cdaa-4107-8567-e8cd9433754b","slug":"mistral-large","name":"Mistral Large","family":"Mistral Large","version":"1","description":"Mistral's first proprietary flagship API model, introduced alongside Le Chat and stronger multilingual/coding performance.","orgSlug":"mistral","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":32000,"status":"deprecated","releasedDate":"2024-02-26","homepageUrl":"https://mistral.ai/news/mistral-large/","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"43d7d3e7-c058-46a2-85ec-4e2b5cd65383","slug":"deepseek-r1-lite-preview","name":"DeepSeek-R1-Lite-Preview","family":"DeepSeek R1","version":"Lite Preview","description":"Reasoning-preview model exposed in DeepSeek Chat ahead of the open DeepSeek-R1 release.","orgSlug":"deepseek","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"retired","releasedDate":"2024-11-20","homepageUrl":"https://api-docs.deepseek.com/news/news1120","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"d399ee57-b2d1-4c3a-8d42-96e2c862bd16","slug":"deepseek-v2-5","name":"DeepSeek-V2.5","family":"DeepSeek V2","version":"2.5","description":"Unified DeepSeek V2 generation combining general-chat and coding strengths before the V3 series.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"DeepSeek License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":236,"paramsActiveB":21,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-09-05","homepageUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V2.5","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V2.5","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"decb0063-aebe-4e5f-811c-9348c146aa63","slug":"deepseek-coder-v2","name":"DeepSeek-Coder-V2","family":"DeepSeek Coder","version":"V2","description":"Open code-focused MoE built from DeepSeek-V2, expanding programming-language coverage and coding benchmark performance.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"DeepSeek License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":236,"paramsActiveB":21,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-06-17","homepageUrl":"https://github.com/deepseek-ai/DeepSeek-Coder-V2","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"f10a416b-9e53-4869-bdf3-e37058ad083f","slug":"deepseek-v2","name":"DeepSeek-V2","family":"DeepSeek V2","version":"2","description":"DeepSeek's first major MoE general model with Multi-head Latent Attention and low-cost API positioning.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"DeepSeek License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":236,"paramsActiveB":21,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-05-07","homepageUrl":"https://github.com/deepseek-ai/DeepSeek-V2","modelCardUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V2","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"f9a399c2-ea87-4bd0-a65f-d2811ec22b99","slug":"deepseek-moe-16b","name":"DeepSeekMoE 16B","family":"DeepSeekMoE","version":"16B","description":"Early DeepSeek sparse MoE research model that foreshadowed the later V2/V3 architecture direction.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"DeepSeek License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":16,"paramsActiveB":2.8,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2024-01-11","homepageUrl":"https://github.com/deepseek-ai/DeepSeek-MoE","modelCardUrl":"https://huggingface.co/deepseek-ai/deepseek-moe-16b-chat","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"548186e0-251c-4d4e-98eb-7503382bef3b","slug":"deepseek-llm-67b","name":"DeepSeek LLM 67B","family":"DeepSeek LLM","version":"67B","description":"First general DeepSeek language model family, with 7B and 67B base/chat checkpoints.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"DeepSeek License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":67,"paramsActiveB":67,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2023-11-29","homepageUrl":"https://github.com/deepseek-ai/DeepSeek-LLM","modelCardUrl":"https://huggingface.co/deepseek-ai/deepseek-llm-67b-chat","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"f84ebfd3-5bc1-47d9-b8e1-45302434047e","slug":"deepseek-coder-33b","name":"DeepSeek Coder 33B","family":"DeepSeek Coder","version":"33B","description":"DeepSeek's first public code-model family, released before the general DeepSeek LLM line.","orgSlug":"deepseek","isFrontier":false,"licenseType":"open_source","licenseName":"DeepSeek License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":33,"paramsActiveB":33,"isSizeDisclosed":true,"contextWindow":16000,"status":"available","releasedDate":"2023-11-02","homepageUrl":"https://github.com/deepseek-ai/DeepSeek-Coder","modelCardUrl":"https://huggingface.co/deepseek-ai/deepseek-coder-33b-instruct","notable":false,"org":{"id":"de082675-89ad-43e2-840f-4991588b30d8","slug":"deepseek","name":"DeepSeek","shortName":"DeepSeek","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"edb2c9d7-f065-452b-adba-cc41a81fd3fd","slug":"qwen3-235b-a22b","name":"Qwen3-235B-A22B","family":"Qwen3","version":"235B-A22B","description":"Largest open Qwen3 MoE, introducing hybrid thinking/non-thinking modes and 119-language coverage.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":235,"paramsActiveB":22,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-04-28","homepageUrl":"https://qwenlm.github.io/blog/qwen3/","modelCardUrl":"https://huggingface.co/Qwen/Qwen3-235B-A22B","notable":true,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"1e6fa358-63de-474c-882d-5470b21da1a8","slug":"qwen2-5-vl-72b","name":"Qwen2.5-VL-72B","family":"Qwen2.5-VL","version":"72B","description":"Vision-language Qwen2.5 model for image, document, video, and agentic visual grounding tasks.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Qwen License","weightsAvailable":true,"modalities":["text","vision","video"],"architecture":"dense","paramsTotalB":72,"paramsActiveB":72,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-01-26","homepageUrl":"https://qwenlm.github.io/blog/qwen2.5-vl/","modelCardUrl":"https://huggingface.co/Qwen/Qwen2.5-VL-72B-Instruct","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"a5afb5ee-bd21-4825-aba9-0983650dca59","slug":"qwen2-5-max","name":"Qwen2.5-Max","family":"Qwen2.5","version":"Max","description":"Proprietary MoE flagship for the Qwen2.5 generation, released through Qwen Chat and Alibaba Cloud APIs.","orgSlug":"alibaba","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"moe","isSizeDisclosed":false,"status":"available","releasedDate":"2025-01-29","homepageUrl":"https://qwenlm.github.io/blog/qwen2.5-max/","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"81f6a6ce-7e7e-4550-8ae9-559ba0cbdb56","slug":"qwq-32b-preview","name":"QwQ-32B-Preview","family":"QwQ","version":"32B Preview","description":"Qwen's first public reasoning-preview model, aimed at math, coding, and deliberate problem solving.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":32,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2024-11-28","homepageUrl":"https://qwenlm.github.io/blog/qwq-32b-preview/","modelCardUrl":"https://huggingface.co/Qwen/QwQ-32B-Preview","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"11a15c79-3f44-486f-a411-85c5b6f951c8","slug":"qwen2-5-coder-32b","name":"Qwen2.5-Coder-32B","family":"Qwen2.5-Coder","version":"32B","description":"Code-specialized Qwen2.5 model family, with the 32B checkpoint as the flagship open coding model.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":32,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-11-12","homepageUrl":"https://qwenlm.github.io/blog/qwen2.5-coder-family/","modelCardUrl":"https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"59d6751b-126f-44e5-a866-bdcc3516d734","slug":"qwen2-5-72b","name":"Qwen2.5-72B","family":"Qwen2.5","version":"72B","description":"Broad Qwen2.5 foundation-model update spanning general, coding, math, and multimodal descendants.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Qwen License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":72,"paramsActiveB":72,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-09-19","homepageUrl":"https://qwenlm.github.io/blog/qwen2.5/","modelCardUrl":"https://huggingface.co/Qwen/Qwen2.5-72B-Instruct","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"7b7a7f49-5982-4fda-a711-7d9923917257","slug":"qwen2-72b","name":"Qwen2-72B","family":"Qwen2","version":"72B","description":"Qwen2's largest dense model, introducing stronger multilingual support, coding/math gains, and long-context variants.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Qwen License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":72,"paramsActiveB":72,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-06-07","homepageUrl":"https://qwenlm.github.io/blog/qwen2/","modelCardUrl":"https://huggingface.co/Qwen/Qwen2-72B-Instruct","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"35c0f594-f1a1-4300-bbd3-c24a57c41be6","slug":"qwen1-5-110b","name":"Qwen1.5-110B","family":"Qwen1.5","version":"110B","description":"Largest Qwen1.5 model, released as the bridge from the original Qwen line to Qwen2.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Tongyi Qianwen License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":110,"paramsActiveB":110,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2024-02-05","homepageUrl":"https://qwenlm.github.io/blog/qwen1.5/","modelCardUrl":"https://huggingface.co/Qwen/Qwen1.5-110B-Chat","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"5d1e0162-5469-46a5-a5cd-e914dba97492","slug":"qwen-14b","name":"Qwen-14B","family":"Qwen","version":"14B","description":"Second open Qwen size, expanding the first-generation Qwen language-model lineup.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Tongyi Qianwen License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":14,"paramsActiveB":14,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2023-09-25","homepageUrl":"https://qwenlm.github.io/blog/qwen/","modelCardUrl":"https://huggingface.co/Qwen/Qwen-14B-Chat","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"e854506c-b689-412c-a902-83164c5cf827","slug":"qwen-7b","name":"Qwen-7B","family":"Qwen","version":"7B","description":"Alibaba's first open Qwen checkpoint and the start of the Qwen open-model line.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Tongyi Qianwen License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2023-08-03","homepageUrl":"https://qwenlm.github.io/blog/qwen/","modelCardUrl":"https://huggingface.co/Qwen/Qwen-7B-Chat","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"c327d9bb-0bb6-4aba-946f-0ec951623727","slug":"kimi-1m","name":"Kimi 1M","family":"Kimi","version":"1M","description":"Long-context Kimi upgrade advertised with support for million-character document and conversation contexts.","orgSlug":"moonshot","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-03-18","homepageUrl":"https://kimi.moonshot.cn/","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"5a8dbf84-b46e-45b4-b3e9-34789ada8133","slug":"kimi-chat","name":"Kimi Chat","family":"Kimi","version":"Chat","description":"Moonshot's first Kimi assistant release, establishing the long-context product line before the open Kimi model cards.","orgSlug":"moonshot","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2023-10-09","homepageUrl":"https://kimi.moonshot.cn/","notable":false,"org":{"id":"c55adbbe-5911-44c1-8285-676371328dd4","slug":"moonshot","name":"Moonshot AI","shortName":"Moonshot","orgType":"open_weight_lab","countryCode":"CN","countryName":"China"}},{"id":"b904f4e1-dbb3-4191-a050-65b5148214bb","slug":"aya-23-35b","name":"Aya 23 35B","family":"Aya","version":"23 35B","description":"Open multilingual research model covering 23 languages, released by Cohere For AI.","orgSlug":"cohere","isFrontier":false,"licenseType":"open_weight","licenseName":"CC-BY-NC-4.0","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":35,"paramsActiveB":35,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-05-23","homepageUrl":"https://cohere.com/research/aya","modelCardUrl":"https://huggingface.co/CohereForAI/aya-23-35B","notable":false,"org":{"id":"17f4e087-94d7-4373-a3c2-8d365d1e0e16","slug":"cohere","name":"Cohere","shortName":"Cohere","orgType":"startup","countryCode":"CA","countryName":"Canada"}},{"id":"8b2b4a3a-9952-4e4d-980b-abbd83d02190","slug":"command-r-plus","name":"Command R+","family":"Command","version":"R+","description":"Higher-capability RAG and tool-use model in Cohere's Command R family.","orgSlug":"cohere","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":128000,"status":"deprecated","releasedDate":"2024-04-04","homepageUrl":"https://cohere.com/blog/command-r-plus","notable":false,"org":{"id":"17f4e087-94d7-4373-a3c2-8d365d1e0e16","slug":"cohere","name":"Cohere","shortName":"Cohere","orgType":"startup","countryCode":"CA","countryName":"Canada"}},{"id":"aa9efa06-2626-4208-85e4-3db2a97d35bd","slug":"command-r","name":"Command R","family":"Command","version":"R","description":"Enterprise RAG-focused model with tool use, citations, multilingual retrieval, and long-context support.","orgSlug":"cohere","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":128000,"status":"deprecated","releasedDate":"2024-03-11","homepageUrl":"https://cohere.com/blog/command-r","notable":false,"org":{"id":"17f4e087-94d7-4373-a3c2-8d365d1e0e16","slug":"cohere","name":"Cohere","shortName":"Cohere","orgType":"startup","countryCode":"CA","countryName":"Canada"}},{"id":"aab25611-d020-48dc-bcfb-0779193aa5bf","slug":"phi-4-reasoning","name":"Phi-4 Reasoning","family":"Phi","version":"4 Reasoning","description":"Phi-4 reasoning-specialized model family for math, science, and chain-of-thought style tasks.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_weight","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":14,"paramsActiveB":14,"isSizeDisclosed":true,"status":"available","releasedDate":"2025-04-30","homepageUrl":"https://huggingface.co/microsoft/Phi-4-reasoning","modelCardUrl":"https://huggingface.co/microsoft/Phi-4-reasoning","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"af6c0faf-37dc-4498-bd97-eb16413190ce","slug":"phi-3-5-moe","name":"Phi-3.5 MoE","family":"Phi","version":"3.5 MoE","description":"Phi-3.5 mixture-of-experts model, scaling Microsoft's small-model line while preserving efficient active parameters.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_weight","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":42,"paramsActiveB":6.6,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-08-20","homepageUrl":"https://huggingface.co/microsoft/Phi-3.5-MoE-instruct","modelCardUrl":"https://huggingface.co/microsoft/Phi-3.5-MoE-instruct","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"95d3bbd1-55a4-43ec-b6c8-2811efda8d31","slug":"phi-3-mini","name":"Phi-3 Mini","family":"Phi","version":"3 Mini","description":"3.8B-parameter Phi-3 model released as a phone-capable small model with 4K and 128K variants.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_weight","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":3.8,"paramsActiveB":3.8,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-04-23","homepageUrl":"https://www.microsoft.com/en-us/research/publication/phi-3-technical-report-a-highly-capable-language-model-locally-on-your-phone/","modelCardUrl":"https://huggingface.co/microsoft/Phi-3-mini-128k-instruct","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"90cd09da-ec44-445b-be27-33c33f9a2dcf","slug":"phi-2","name":"Phi-2","family":"Phi","version":"2","description":"2.7B-parameter Phi model showing strong reasoning and language understanding at small scale.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_weight","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":2.7,"paramsActiveB":2.7,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-12-12","homepageUrl":"https://www.microsoft.com/en-us/research/blog/phi-2-the-surprising-power-of-small-language-models/","modelCardUrl":"https://huggingface.co/microsoft/phi-2","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"92c3f7d2-1535-4d59-95b6-729cf3d37e0b","slug":"phi-1","name":"Phi-1","family":"Phi","version":"1","description":"Microsoft's first Phi small-language-model release, demonstrating strong code performance from textbook-quality synthetic data.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_weight","licenseName":"MIT","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":1.3,"paramsActiveB":1.3,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-06-21","homepageUrl":"https://www.microsoft.com/en-us/research/blog/textbooks-are-all-you-need/","modelCardUrl":"https://huggingface.co/microsoft/phi-1","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"767c85e7-6dbe-48b4-83d1-2e05ee95e83f","slug":"amazon-nova-lite","name":"Amazon Nova Lite","family":"Amazon Nova","version":"Lite","description":"Lower-cost multimodal Nova understanding model for text, image, and video inputs.","orgSlug":"amazon","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","video"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":300000,"status":"available","releasedDate":"2024-12-03","homepageUrl":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","notable":false,"org":{"id":"4daa4336-3d5e-42df-a61d-a6efad4c756d","slug":"amazon","name":"Amazon","shortName":"Amazon","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"a8bcce44-cd0a-4328-a037-d557df99812a","slug":"amazon-titan-text-premier","name":"Amazon Titan Text Premier","family":"Amazon Titan","version":"Text Premier","description":"Larger Titan text model for enterprise RAG, summarization, and agent workflows in Amazon Bedrock.","orgSlug":"amazon","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-04-30","homepageUrl":"https://aws.amazon.com/blogs/aws/amazon-titan-text-premier-is-now-available-in-amazon-bedrock/","notable":false,"org":{"id":"4daa4336-3d5e-42df-a61d-a6efad4c756d","slug":"amazon","name":"Amazon","shortName":"Amazon","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"6a3f2020-47bc-49af-a17d-93606aad88c2","slug":"amazon-titan-text-express","name":"Amazon Titan Text Express","family":"Amazon Titan","version":"Text Express","description":"Amazon's first-party Titan text generation model exposed through Bedrock, initially alongside embeddings and image models.","orgSlug":"amazon","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2023-09-28","homepageUrl":"https://aws.amazon.com/blogs/aws/amazon-bedrock-is-now-generally-available-build-and-scale-generative-ai-applications-with-foundation-models/","notable":false,"org":{"id":"4daa4336-3d5e-42df-a61d-a6efad4c756d","slug":"amazon","name":"Amazon","shortName":"Amazon","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"8774c5fd-9d96-4365-893b-1ece3ce33be3","slug":"granite-3-3-8b","name":"Granite 3.3 8B","family":"Granite 3","version":"3.3 8B","description":"Granite 3.3 text update for enterprise chat, RAG, and instruction-following workflows.","orgSlug":"ibm","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-04-30","homepageUrl":"https://huggingface.co/ibm-granite/granite-3.3-8b-instruct","modelCardUrl":"https://huggingface.co/ibm-granite/granite-3.3-8b-instruct","notable":false,"org":{"id":"313d3992-f81b-4751-8793-51c125444f42","slug":"ibm","name":"IBM","shortName":"IBM","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"7585e8db-ed13-410f-923f-4ee6ca79bedf","slug":"granite-3-2-8b","name":"Granite 3.2 8B","family":"Granite 3","version":"3.2 8B","description":"Granite 3.2 update with reasoning controls and multimodal/document-oriented Granite variants.","orgSlug":"ibm","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-02-26","homepageUrl":"https://www.ibm.com/new/announcements/ibm-granite-3-2-expands-enterprise-ai-options","modelCardUrl":"https://huggingface.co/ibm-granite/granite-3.2-8b-instruct","notable":false,"org":{"id":"313d3992-f81b-4751-8793-51c125444f42","slug":"ibm","name":"IBM","shortName":"IBM","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"6bfe3874-4815-4256-9ce1-cd8e7da17c10","slug":"granite-3-0-8b","name":"Granite 3.0 8B","family":"Granite 3","version":"3.0 8B","description":"Apache-licensed Granite 3.0 text model, part of IBM's push toward enterprise-friendly open models.","orgSlug":"ibm","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2024-10-21","homepageUrl":"https://www.ibm.com/new/announcements/ibm-granite-3-0-open-state-of-the-art-enterprise-models","modelCardUrl":"https://huggingface.co/ibm-granite/granite-3.0-8b-instruct","notable":false,"org":{"id":"313d3992-f81b-4751-8793-51c125444f42","slug":"ibm","name":"IBM","shortName":"IBM","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"d0aed805-fd1d-4876-a767-38198bb433ce","slug":"granite-13b","name":"Granite 13B","family":"Granite","version":"13B","description":"IBM's early Granite foundation model family for enterprise language and code tasks.","orgSlug":"ibm","isFrontier":false,"licenseType":"open_weight","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":13,"paramsActiveB":13,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-09-07","homepageUrl":"https://www.ibm.com/granite","modelCardUrl":"https://huggingface.co/ibm-granite/granite-13b-chat-v2","notable":false,"org":{"id":"313d3992-f81b-4751-8793-51c125444f42","slug":"ibm","name":"IBM","shortName":"IBM","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"e5d42fd4-319b-4d46-b1d5-24186e9b7318","slug":"tulu-3-405b","name":"Tulu 3 405B","family":"Tulu 3","version":"405B","description":"Ai2's post-trained open instruction model line, scaling the Tulu recipe to Llama 3.1 405B.","orgSlug":"allen-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.1 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":405,"paramsActiveB":405,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-11-21","homepageUrl":"https://allenai.org/blog/tulu-3","modelCardUrl":"https://huggingface.co/allenai/Llama-3.1-Tulu-3-405B","notable":false,"org":{"id":"d38a5c02-ed52-4f11-8d3c-d6b5b507086e","slug":"allen-ai","name":"Allen Institute for AI (Ai2)","shortName":"Ai2","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"f910fdd1-fea4-4fa3-b9c8-1465159eec83","slug":"molmo-72b","name":"Molmo 72B","family":"Molmo","version":"72B","description":"Open multimodal model family trained for strong image understanding, pointing, and visual grounding.","orgSlug":"allen-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"Apache-2.0 / OLMo License","weightsAvailable":true,"modalities":["text","vision"],"architecture":"dense","paramsTotalB":72,"paramsActiveB":72,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-09-25","homepageUrl":"https://molmo.allenai.org/","modelCardUrl":"https://huggingface.co/allenai/Molmo-72B-0924","notable":false,"org":{"id":"d38a5c02-ed52-4f11-8d3c-d6b5b507086e","slug":"allen-ai","name":"Allen Institute for AI (Ai2)","shortName":"Ai2","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"6a20b537-9fa5-4dcf-82e3-18d31e610a2a","slug":"olmoe-1b-7b","name":"OLMoE 1B-7B","family":"OLMoE","version":"1B-7B","description":"Fully open sparse MoE model with 7B total and about 1B active parameters.","orgSlug":"allen-ai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":7,"paramsActiveB":1,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-09-03","homepageUrl":"https://allenai.org/blog/olmoe","modelCardUrl":"https://huggingface.co/allenai/OLMoE-1B-7B-0924-Instruct","notable":false,"org":{"id":"d38a5c02-ed52-4f11-8d3c-d6b5b507086e","slug":"allen-ai","name":"Allen Institute for AI (Ai2)","shortName":"Ai2","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"5d5f4661-423d-4133-afc8-0fe7f852b8a3","slug":"olmo-7b","name":"OLMo 7B","family":"OLMo","version":"7B","description":"Ai2's first fully open language model release, including weights, training data, code, logs, and intermediate checkpoints.","orgSlug":"allen-ai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2024-02-01","homepageUrl":"https://allenai.org/blog/olmo-open-language-model","modelCardUrl":"https://huggingface.co/allenai/OLMo-7B-Instruct","notable":false,"org":{"id":"d38a5c02-ed52-4f11-8d3c-d6b5b507086e","slug":"allen-ai","name":"Allen Institute for AI (Ai2)","shortName":"Ai2","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"f94b5248-5822-4926-a2cb-89df21db09a5","slug":"falcon-2-11b","name":"Falcon 2 11B","family":"Falcon 2","version":"11B","description":"Falcon 2 generation, including text and vision-language 11B models under a permissive TII license.","orgSlug":"tii","isFrontier":false,"licenseType":"open_weight","licenseName":"TII Falcon License","weightsAvailable":true,"modalities":["text","vision"],"architecture":"dense","paramsTotalB":11,"paramsActiveB":11,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2024-05-13","homepageUrl":"https://www.tii.ae/news/technology-innovation-institute-launches-falcon-2-series","modelCardUrl":"https://huggingface.co/tiiuae/falcon-11B","notable":false,"org":{"id":"47a447cd-6251-4c25-991e-2f17f8c71c8b","slug":"tii","name":"Technology Innovation Institute","shortName":"TII","orgType":"research_org","countryCode":"AE","countryName":"United Arab Emirates"}},{"id":"9657ab0d-fbc2-4912-b907-08bd27e540b5","slug":"falcon-40b","name":"Falcon 40B","family":"Falcon","version":"40B","description":"TII's breakout open Falcon model, released before Falcon 180B and trained on the RefinedWeb corpus.","orgSlug":"tii","isFrontier":false,"licenseType":"open_weight","licenseName":"Falcon License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":40,"paramsActiveB":40,"isSizeDisclosed":true,"contextWindow":2048,"status":"available","releasedDate":"2023-05-25","homepageUrl":"https://falconllm.tii.ae/","modelCardUrl":"https://huggingface.co/tiiuae/falcon-40b-instruct","paperUrl":"https://arxiv.org/abs/2311.16867","notable":false,"org":{"id":"47a447cd-6251-4c25-991e-2f17f8c71c8b","slug":"tii","name":"Technology Innovation Institute","shortName":"TII","orgType":"research_org","countryCode":"AE","countryName":"United Arab Emirates"}},{"id":"7df50f86-3d9d-4495-9dd8-05b1d96868fc","slug":"jamba","name":"Jamba","family":"Jamba","version":"1.0","description":"First Jamba hybrid Transformer-Mamba MoE model with open weights and a 256K context length.","orgSlug":"ai21","isFrontier":false,"licenseType":"open_weight","licenseName":"Jamba Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":52,"paramsActiveB":12,"isSizeDisclosed":true,"contextWindow":256000,"status":"available","releasedDate":"2024-03-28","homepageUrl":"https://www.ai21.com/blog/announcing-jamba","modelCardUrl":"https://huggingface.co/ai21labs/Jamba-v0.1","paperUrl":"https://arxiv.org/abs/2403.19887","notable":false,"org":{"id":"789fdb1b-121d-42db-ad8e-c92b3a82fd8f","slug":"ai21","name":"AI21 Labs","shortName":"AI21","orgType":"startup","countryCode":"IL","countryName":"Israel"}},{"id":"ee2bcc72-9a39-4de2-a6f0-f71ef0ce0b99","slug":"jurassic-2-ultra","name":"Jurassic-2 Ultra","family":"Jurassic","version":"2 Ultra","description":"Second-generation Jurassic model with better multilingual support, lower latency, and instruction following.","orgSlug":"ai21","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"status":"deprecated","releasedDate":"2023-03-09","homepageUrl":"https://www.ai21.com/","notable":false,"org":{"id":"789fdb1b-121d-42db-ad8e-c92b3a82fd8f","slug":"ai21","name":"AI21 Labs","shortName":"AI21","orgType":"startup","countryCode":"IL","countryName":"Israel"}},{"id":"2b8c5042-bb98-4663-ac86-e2b3122c6da0","slug":"llama-3-1-nemotron-70b","name":"Llama-3.1-Nemotron-70B","family":"Nemotron","version":"Llama 3.1 70B","description":"NVIDIA-tuned Llama 3.1 70B instruction model optimized with Nemotron reward and alignment recipes.","orgSlug":"nvidia","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.1 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":70,"paramsActiveB":70,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-10-15","homepageUrl":"https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct-HF","modelCardUrl":"https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct-HF","notable":false,"org":{"id":"f32419b4-52f1-4de7-86dd-1d8370296ada","slug":"nvidia","name":"NVIDIA","shortName":"NVIDIA","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"c03b2cb6-8668-4c14-a964-ba1227689364","slug":"nemotron-4-340b","name":"Nemotron-4 340B","family":"Nemotron 4","version":"340B","description":"NVIDIA's large open model family for synthetic data generation and reward modeling.","orgSlug":"nvidia","isFrontier":false,"licenseType":"open_weight","licenseName":"NVIDIA Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":340,"paramsActiveB":340,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2024-06-14","homepageUrl":"https://huggingface.co/nvidia/Nemotron-4-340B-Instruct","modelCardUrl":"https://huggingface.co/nvidia/Nemotron-4-340B-Instruct","notable":false,"org":{"id":"f32419b4-52f1-4de7-86dd-1d8370296ada","slug":"nvidia","name":"NVIDIA","shortName":"NVIDIA","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"87896e87-5a06-42fd-8eec-7817438184e6","slug":"ernie-x1","name":"ERNIE X1","family":"ERNIE","version":"X1","description":"Baidu's reasoning model released alongside ERNIE 4.5 before the open ERNIE 4.5 weights.","orgSlug":"baidu","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2025-03-16","homepageUrl":"https://yiyan.baidu.com/","notable":false,"org":{"id":"2a4d1508-7601-432d-8c75-ea2aa0da12d1","slug":"baidu","name":"Baidu","shortName":"Baidu","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"afee8896-06b7-445f-b275-242706b6125c","slug":"ernie-4-0","name":"ERNIE 4.0","family":"ERNIE","version":"4.0","description":"Baidu's fourth-generation ERNIE flagship, announced with stronger understanding, generation, reasoning, and memory.","orgSlug":"baidu","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2023-10-17","homepageUrl":"https://yiyan.baidu.com/","notable":false,"org":{"id":"2a4d1508-7601-432d-8c75-ea2aa0da12d1","slug":"baidu","name":"Baidu","shortName":"Baidu","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"94c7b1b4-e319-4ea3-8077-2147d33a989c","slug":"ernie-bot","name":"ERNIE Bot","family":"ERNIE","version":"Bot","description":"Baidu's public chat assistant launch, built on the ERNIE foundation-model line.","orgSlug":"baidu","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2023-03-16","homepageUrl":"https://yiyan.baidu.com/","notable":false,"org":{"id":"2a4d1508-7601-432d-8c75-ea2aa0da12d1","slug":"baidu","name":"Baidu","shortName":"Baidu","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"ece9a05e-10c6-47b7-96e5-d8cd5d984e11","slug":"ernie-3-titan","name":"ERNIE 3.0 Titan","family":"ERNIE","version":"3.0 Titan","description":"Baidu's 260B-parameter ERNIE 3.0 Titan model, an early Chinese frontier-scale language model.","orgSlug":"baidu","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"dense","paramsTotalB":260,"paramsActiveB":260,"isSizeDisclosed":true,"status":"retired","releasedDate":"2021-12-08","homepageUrl":"https://arxiv.org/abs/2112.12731","paperUrl":"https://arxiv.org/abs/2112.12731","notable":false,"org":{"id":"2a4d1508-7601-432d-8c75-ea2aa0da12d1","slug":"baidu","name":"Baidu","shortName":"Baidu","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"2cbd0798-c71e-4873-8d0c-92907212d46a","slug":"hunyuan-turbo","name":"Hunyuan Turbo","family":"Hunyuan","version":"Turbo","description":"Tencent's faster, lower-cost Hunyuan update before the open Hunyuan-Large model card.","orgSlug":"tencent","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-09-05","homepageUrl":"https://hunyuan.tencent.com/","notable":false,"org":{"id":"f08b4f2e-ef9e-4ef7-bf7e-be299aded92c","slug":"tencent","name":"Tencent Hunyuan","shortName":"Hunyuan","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"9050c1e0-a9a3-43c4-b320-670c60bcb983","slug":"hunyuan-standard","name":"Hunyuan","family":"Hunyuan","version":"Standard","description":"Tencent's first Hunyuan foundation model release, introduced as a general-purpose Chinese enterprise model.","orgSlug":"tencent","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2023-09-07","homepageUrl":"https://hunyuan.tencent.com/","notable":false,"org":{"id":"f08b4f2e-ef9e-4ef7-bf7e-be299aded92c","slug":"tencent","name":"Tencent Hunyuan","shortName":"Hunyuan","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"b22ea8a6-cbeb-4de5-a5a1-3dbfbe08919f","slug":"seed-thinking-v1-5","name":"Seed Thinking v1.5","family":"Seed Thinking","version":"1.5","description":"ByteDance Seed reasoning model focused on long-horizon thinking and problem solving.","orgSlug":"bytedance","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2025-05-22","homepageUrl":"https://seed.bytedance.com/","notable":false,"org":{"id":"5eada657-5a27-4c3f-868e-a171956035d2","slug":"bytedance","name":"ByteDance Seed","shortName":"Seed","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"9268f803-780a-4859-9ae6-520bcff5ec37","slug":"doubao-1-5-pro","name":"Doubao-1.5-pro","family":"Doubao","version":"1.5 Pro","description":"Doubao 1.5 Pro update positioned for stronger multimodal, reasoning, and agentic work in Volcano Engine.","orgSlug":"bytedance","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2025-01-22","homepageUrl":"https://www.volcengine.com/product/doubao","notable":false,"org":{"id":"5eada657-5a27-4c3f-868e-a171956035d2","slug":"bytedance","name":"ByteDance Seed","shortName":"Seed","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"42b63eaf-d082-471b-81d2-1fd383db7574","slug":"doubao-pro","name":"Doubao-pro","family":"Doubao","version":"Pro","description":"ByteDance's commercial Doubao foundation model line for text, code, and assistant workloads.","orgSlug":"bytedance","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-05-15","homepageUrl":"https://www.volcengine.com/product/doubao","notable":false,"org":{"id":"5eada657-5a27-4c3f-868e-a171956035d2","slug":"bytedance","name":"ByteDance Seed","shortName":"Seed","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"5c9ef2d2-fbcb-4371-a294-ff467a310f50","slug":"minimax-01","name":"MiniMax-01","family":"MiniMax","version":"01","description":"Open MiniMax generation with MiniMax-Text-01 and MiniMax-VL-01 long-context models.","orgSlug":"minimax","isFrontier":false,"licenseType":"open_weight","licenseName":"MiniMax Model License","weightsAvailable":true,"modalities":["text","vision","code"],"architecture":"hybrid","paramsTotalB":456,"paramsActiveB":45.9,"isSizeDisclosed":true,"contextWindow":4000000,"status":"available","releasedDate":"2025-01-15","homepageUrl":"https://huggingface.co/MiniMaxAI/MiniMax-Text-01","modelCardUrl":"https://huggingface.co/MiniMaxAI/MiniMax-Text-01","notable":false,"org":{"id":"1ba738a6-44be-4db6-afa3-59f997431b3b","slug":"minimax","name":"MiniMax","shortName":"MiniMax","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"cf3ffd1f-997b-4ecb-8b65-9d4661e531ae","slug":"minimax-abab6-5","name":"abab6.5","family":"abab","version":"6.5","description":"MiniMax's commercial long-context abab model generation before the open MiniMax-01 and M series.","orgSlug":"minimax","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text"],"architecture":"unknown","isSizeDisclosed":false,"contextWindow":1000000,"status":"available","releasedDate":"2024-04-17","homepageUrl":"https://www.minimax.io/","notable":false,"org":{"id":"1ba738a6-44be-4db6-afa3-59f997431b3b","slug":"minimax","name":"MiniMax","shortName":"MiniMax","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"fa20327f-596e-48a8-ab23-3b84f39824d7","slug":"step-2","name":"Step-2","family":"Step","version":"2","description":"Second-generation StepFun foundation model line with larger-scale multimodal and reasoning ambitions.","orgSlug":"stepfun","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-12-23","homepageUrl":"https://www.stepfun.com/","notable":false,"org":{"id":"aa1c8268-616e-4866-a5eb-542af5011ec4","slug":"stepfun","name":"StepFun","shortName":"StepFun","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"88d26d66-b503-440e-98ef-5b16783c1b3e","slug":"step-1v","name":"Step-1V","family":"Step","version":"1V","description":"StepFun's first major vision-language model, released after the Step-1 language model.","orgSlug":"stepfun","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-04-12","homepageUrl":"https://www.stepfun.com/","notable":false,"org":{"id":"aa1c8268-616e-4866-a5eb-542af5011ec4","slug":"stepfun","name":"StepFun","shortName":"StepFun","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"8519e3c3-5c1d-4507-8317-915a37c9c64a","slug":"step-1","name":"Step-1","family":"Step","version":"1","description":"StepFun's first public foundation model generation, introduced as a trillion-parameter Chinese model line.","orgSlug":"stepfun","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","code"],"architecture":"unknown","isSizeDisclosed":false,"status":"available","releasedDate":"2024-03-23","homepageUrl":"https://www.stepfun.com/","notable":false,"org":{"id":"aa1c8268-616e-4866-a5eb-542af5011ec4","slug":"stepfun","name":"StepFun","shortName":"StepFun","orgType":"startup","countryCode":"CN","countryName":"China"}},{"id":"82124550-c89a-49ec-a60b-3437cab83eb7","slug":"exaone-3-5-32b","name":"EXAONE 3.5 32B","family":"EXAONE 3.5","version":"32B","description":"EXAONE 3.5 32B open-weight model for bilingual reasoning, coding, and long-context tasks.","orgSlug":"lg-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"EXAONE AI Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":32,"paramsActiveB":32,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2024-12-09","homepageUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-3.5-32B-Instruct","modelCardUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-3.5-32B-Instruct","notable":false,"org":{"id":"9eee4d26-483e-40c3-8351-3b54653c5294","slug":"lg-ai","name":"LG AI Research","shortName":"LG AI","orgType":"research_org","countryCode":"KR","countryName":"South Korea"}},{"id":"910cd9bb-8361-4df2-b727-51d4645fd929","slug":"exaone-3-0-7-8b","name":"EXAONE 3.0 7.8B","family":"EXAONE 3","version":"3.0 7.8B","description":"LG's first open-weight EXAONE model, a compact bilingual instruction model for Korean and English.","orgSlug":"lg-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"EXAONE AI Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7.8,"paramsActiveB":7.8,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-08-07","homepageUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct","modelCardUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct","notable":false,"org":{"id":"9eee4d26-483e-40c3-8351-3b54653c5294","slug":"lg-ai","name":"LG AI Research","shortName":"LG AI","orgType":"research_org","countryCode":"KR","countryName":"South Korea"}},{"id":"da2191f7-527f-47d5-9b1a-35991eba0477","slug":"exaone-2","name":"EXAONE 2.0","family":"EXAONE","version":"2.0","description":"Second EXAONE generation, improving bilingual Korean-English performance and enterprise deployment options.","orgSlug":"lg-ai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision"],"architecture":"unknown","isSizeDisclosed":false,"status":"retired","releasedDate":"2023-07-19","homepageUrl":"https://www.lgresearch.ai/","notable":false,"org":{"id":"9eee4d26-483e-40c3-8351-3b54653c5294","slug":"lg-ai","name":"LG AI Research","shortName":"LG AI","orgType":"research_org","countryCode":"KR","countryName":"South Korea"}},{"id":"f13f10e5-4694-4805-a602-55b1ffd63c7a","slug":"exaone-1","name":"EXAONE 1.0","family":"EXAONE","version":"1.0","description":"LG AI Research's first EXAONE foundation model generation, introduced as a large multimodal expert AI.","orgSlug":"lg-ai","isFrontier":false,"licenseType":"proprietary","weightsAvailable":false,"modalities":["text","vision"],"architecture":"unknown","isSizeDisclosed":false,"status":"retired","releasedDate":"2021-12-14","homepageUrl":"https://www.lgresearch.ai/","notable":false,"org":{"id":"9eee4d26-483e-40c3-8351-3b54653c5294","slug":"lg-ai","name":"LG AI Research","shortName":"LG AI","orgType":"research_org","countryCode":"KR","countryName":"South Korea"}},{"id":"cedb93a9-63c3-48be-ab5f-98cefcc76d95","slug":"deephermes-3-llama-3-8b","name":"DeepHermes 3 Llama 3 8B","family":"DeepHermes 3","version":"Llama 3 8B","description":"Nous reasoning-oriented Hermes model trained to combine concise answers with optional deep reasoning traces.","orgSlug":"nous","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2025-02-18","homepageUrl":"https://huggingface.co/NousResearch/DeepHermes-3-Llama-3-8B-Preview","modelCardUrl":"https://huggingface.co/NousResearch/DeepHermes-3-Llama-3-8B-Preview","notable":false,"org":{"id":"b42df71d-a2f8-4df6-bd7b-5a8982031d20","slug":"nous","name":"Nous Research","shortName":"Nous","orgType":"open_weight_lab","countryCode":"US","countryName":"United States"}},{"id":"a0b21659-17b1-4da7-a0ab-b5d3b47038ff","slug":"hermes-3-llama-3-1-405b","name":"Hermes 3 Llama 3.1 405B","family":"Hermes 3","version":"Llama 3.1 405B","description":"Large Hermes 3 instruction-tuned model built on Meta's Llama 3.1 405B.","orgSlug":"nous","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.1 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":405,"paramsActiveB":405,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2024-08-15","homepageUrl":"https://huggingface.co/NousResearch/Hermes-3-Llama-3.1-405B","modelCardUrl":"https://huggingface.co/NousResearch/Hermes-3-Llama-3.1-405B","notable":false,"org":{"id":"b42df71d-a2f8-4df6-bd7b-5a8982031d20","slug":"nous","name":"Nous Research","shortName":"Nous","orgType":"open_weight_lab","countryCode":"US","countryName":"United States"}},{"id":"27860425-84d1-42e7-b2ef-fc0f15524dbf","slug":"nous-hermes-2-mixtral","name":"Nous Hermes 2 Mixtral","family":"Nous Hermes 2","version":"Mixtral 8x7B","description":"Nous instruction-tuned Mixtral model with strong open-chat and tool-use adoption.","orgSlug":"nous","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":47,"paramsActiveB":13,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2024-01-11","homepageUrl":"https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO","modelCardUrl":"https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO","notable":false,"org":{"id":"b42df71d-a2f8-4df6-bd7b-5a8982031d20","slug":"nous","name":"Nous Research","shortName":"Nous","orgType":"open_weight_lab","countryCode":"US","countryName":"United States"}},{"id":"33c552f8-d92d-4569-b67e-92c313bff11e","slug":"nous-hermes-llama2-13b","name":"Nous-Hermes-Llama2-13B","family":"Nous Hermes","version":"Llama2 13B","description":"Early Nous Hermes instruction model on Llama 2, widely used in the open-model fine-tuning ecosystem.","orgSlug":"nous","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 2 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":13,"paramsActiveB":13,"isSizeDisclosed":true,"contextWindow":4096,"status":"available","releasedDate":"2023-07-24","homepageUrl":"https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b","modelCardUrl":"https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b","notable":false,"org":{"id":"b42df71d-a2f8-4df6-bd7b-5a8982031d20","slug":"nous","name":"Nous Research","shortName":"Nous","orgType":"open_weight_lab","countryCode":"US","countryName":"United States"}},{"id":"a219f134-3e56-4af7-80a6-77f7281eb6ed","slug":"sarvam-m","name":"Sarvam-M","family":"Sarvam","version":"M","description":"Sarvam's medium-scale open model for multilingual Indian-language chat, reasoning, and translation tasks.","orgSlug":"sarvam","isFrontier":false,"licenseType":"open_weight","licenseName":"Sarvam AI License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","isSizeDisclosed":false,"status":"available","releasedDate":"2025-05-21","homepageUrl":"https://www.sarvam.ai/blogs/sarvam-m","modelCardUrl":"https://huggingface.co/sarvamai/sarvam-m","notable":false,"org":{"id":"271c4a05-6c35-466d-8b07-9531196e96ab","slug":"sarvam","name":"Sarvam AI","shortName":"Sarvam","orgType":"startup","countryCode":"IN","countryName":"India"}},{"id":"a7922576-c9b0-4b7f-b502-17f63d5e9b3f","slug":"sarvam-1","name":"Sarvam-1","family":"Sarvam","version":"1","description":"Sarvam's 2B open model trained for ten major Indian languages.","orgSlug":"sarvam","isFrontier":false,"licenseType":"open_weight","licenseName":"Sarvam AI License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":2,"paramsActiveB":2,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-10-22","homepageUrl":"https://www.sarvam.ai/blogs/sarvam-1","modelCardUrl":"https://huggingface.co/sarvamai/sarvam-1","notable":false,"org":{"id":"271c4a05-6c35-466d-8b07-9531196e96ab","slug":"sarvam","name":"Sarvam AI","shortName":"Sarvam","orgType":"startup","countryCode":"IN","countryName":"India"}},{"id":"8628624a-676d-46b9-a7a3-4df37d2ac316","slug":"openhathi-7b","name":"OpenHathi-7B","family":"OpenHathi","version":"7B","description":"Sarvam AI's first open Indic language model, adapted from Llama 2 for Hindi and Indian-language work.","orgSlug":"sarvam","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 2 Community License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-12-12","homepageUrl":"https://www.sarvam.ai/blogs/announcing-openhathi","modelCardUrl":"https://huggingface.co/sarvamai/OpenHathi-7B-Hi-v0.1-Base","notable":false,"org":{"id":"271c4a05-6c35-466d-8b07-9531196e96ab","slug":"sarvam","name":"Sarvam AI","shortName":"Sarvam","orgType":"startup","countryCode":"IN","countryName":"India"}},{"id":"5c4fd3f4-5432-431f-88f4-eb3dcbe2fa69","slug":"gpt-oss-20b","name":"gpt-oss-20b","family":"gpt-oss","version":"20B","description":"Smaller gpt-oss reasoning model optimized for local inference on systems with about 16GB of memory.","orgSlug":"openai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":21,"paramsActiveB":3.6,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-08-05","homepageUrl":"https://openai.com/index/introducing-gpt-oss/","modelCardUrl":"https://huggingface.co/openai/gpt-oss-20b","paperUrl":"https://arxiv.org/abs/2508.10925","notable":true,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"7c9d96f1-a0a6-444e-b820-e9ae88da6cb0","slug":"gpt-oss-120b","name":"gpt-oss-120b","family":"gpt-oss","version":"120B","description":"OpenAI's larger open-weight reasoning model, a 117B-total / 5.1B-active MoE with 128K context for local and self-hosted deployment.","orgSlug":"openai","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":117,"paramsActiveB":5.1,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-08-05","homepageUrl":"https://openai.com/index/introducing-gpt-oss/","modelCardUrl":"https://huggingface.co/openai/gpt-oss-120b","paperUrl":"https://arxiv.org/abs/2508.10925","notable":true,"org":{"id":"2c186239-9561-401a-b7f2-917a9a85235e","slug":"openai","name":"OpenAI","shortName":"OpenAI","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"20c4ae6c-ec08-49db-b333-6d917f3cc3bf","slug":"codegemma-7b","name":"CodeGemma 7B","family":"CodeGemma","version":"7B","description":"Open code-specialized Gemma model for local code completion, generation, and instruction-following.","orgSlug":"google-deepmind","isFrontier":false,"licenseType":"open_weight","licenseName":"Gemma Terms of Use","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2024-04-09","homepageUrl":"https://developers.googleblog.com/en/codegemma-open-code-models/","modelCardUrl":"https://huggingface.co/google/codegemma-7b-it","notable":false,"org":{"id":"b0b3712d-cfb9-4a27-9284-b61d83e946db","slug":"google-deepmind","name":"Google DeepMind","shortName":"DeepMind","orgType":"frontier_lab","countryCode":"US","countryName":"United States"}},{"id":"a7ec21ba-54df-4ff9-96c8-ac5904bf1db7","slug":"magistral-small","name":"Magistral Small","family":"Magistral","version":"Small","description":"Open-weight 24B reasoning model from Mistral's Magistral family, popular for local reasoning experiments.","orgSlug":"mistral","isFrontier":false,"licenseType":"open_weight","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":24,"paramsActiveB":24,"isSizeDisclosed":true,"contextWindow":40000,"status":"available","releasedDate":"2025-06-10","homepageUrl":"https://mistral.ai/news/magistral/","modelCardUrl":"https://huggingface.co/mistralai/Magistral-Small-2506","notable":false,"org":{"id":"0ec8eefe-3f88-437c-9752-3a272e1bef3d","slug":"mistral","name":"Mistral AI","shortName":"Mistral","orgType":"open_weight_lab","countryCode":"FR","countryName":"France"}},{"id":"c23a0c0a-583d-45cc-9d1f-49a346a2795b","slug":"qwen2-5-omni-7b","name":"Qwen2.5-Omni-7B","family":"Qwen2.5-Omni","version":"7B","description":"Local omni-modal Qwen model that supports text, image, audio, video, and speech generation in a 7B package.","orgSlug":"alibaba","isFrontier":false,"licenseType":"open_weight","licenseName":"Qwen License","weightsAvailable":true,"modalities":["text","vision","audio","video"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"status":"available","releasedDate":"2025-03-26","homepageUrl":"https://qwenlm.github.io/blog/qwen2.5-omni/","modelCardUrl":"https://huggingface.co/Qwen/Qwen2.5-Omni-7B","notable":false,"org":{"id":"d5cc985c-0fe0-40a0-b13e-1c0d09604b79","slug":"alibaba","name":"Alibaba (Qwen)","shortName":"Qwen","orgType":"big_tech","countryCode":"CN","countryName":"China"}},{"id":"7c6044b4-0df7-43f0-a135-0484e24a81a3","slug":"wizardlm-2-8x22b","name":"WizardLM-2 8x22B","family":"WizardLM","version":"2 8x22B","description":"Microsoft's WizardLM-2 MoE chat model, widely mirrored and run locally after its model-card release.","orgSlug":"microsoft","isFrontier":false,"licenseType":"open_weight","licenseName":"WizardLM-2 License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":141,"paramsActiveB":39,"isSizeDisclosed":true,"contextWindow":65536,"status":"available","releasedDate":"2024-04-15","homepageUrl":"https://huggingface.co/microsoft/WizardLM-2-8x22B","modelCardUrl":"https://huggingface.co/microsoft/WizardLM-2-8x22B","notable":false,"org":{"id":"bafccc3c-5046-4226-b8fe-d5938685d9c9","slug":"microsoft","name":"Microsoft","shortName":"Microsoft","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"82cc933c-7fb9-4adb-8c4b-d6c963bc31e9","slug":"granite-code-34b","name":"Granite Code 34B","family":"Granite Code","version":"34B","description":"Apache-2.0 code model from IBM's Granite Code family, used for local code generation and enterprise coding assistants.","orgSlug":"ibm","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":34,"paramsActiveB":34,"isSizeDisclosed":true,"contextWindow":8192,"status":"available","releasedDate":"2024-05-06","homepageUrl":"https://huggingface.co/ibm-granite/granite-34b-code-instruct","modelCardUrl":"https://huggingface.co/ibm-granite/granite-34b-code-instruct","notable":false,"org":{"id":"313d3992-f81b-4751-8793-51c125444f42","slug":"ibm","name":"IBM","shortName":"IBM","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"8014cf73-0d92-4dc1-8c9c-19b94df71bc6","slug":"smollm2-1-7b","name":"SmolLM2 1.7B","family":"SmolLM","version":"2 1.7B","description":"Compact on-device model family trained on 11T tokens, popular for lightweight local chat and experimentation.","orgSlug":"huggingface","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":1.7,"paramsActiveB":1.7,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-11-04","homepageUrl":"https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct","modelCardUrl":"https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct","paperUrl":"https://arxiv.org/abs/2502.02737","notable":false,"org":{"id":"93268594-f4e0-474f-bca9-53a03d550130","slug":"huggingface","name":"Hugging Face","shortName":"Hugging Face","orgType":"startup","countryCode":"US","countryName":"United States"}},{"id":"97c51964-58de-490b-ae81-4bea04a9a52e","slug":"smollm3-3b","name":"SmolLM3 3B","family":"SmolLM","version":"3 3B","description":"Hugging Face's fully open 3B multilingual long-context model with optional reasoning mode and 128K context.","orgSlug":"huggingface","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":3,"paramsActiveB":3,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-07-08","homepageUrl":"https://huggingface.co/blog/smollm3","modelCardUrl":"https://huggingface.co/HuggingFaceTB/SmolLM3-3B","notable":true,"org":{"id":"93268594-f4e0-474f-bca9-53a03d550130","slug":"huggingface","name":"Hugging Face","shortName":"Hugging Face","orgType":"startup","countryCode":"US","countryName":"United States"}},{"id":"2be4f741-81d0-42a7-813c-0d78f7530ecf","slug":"starcoder2-15b","name":"StarCoder2 15B","family":"StarCoder2","version":"15B","description":"Next-generation BigCode code model trained on 4T+ tokens and 600+ programming languages, with 16K context.","orgSlug":"bigcode","isFrontier":false,"licenseType":"open_weight","licenseName":"BigCode OpenRAIL-M v1","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":16,"paramsActiveB":16,"isSizeDisclosed":true,"contextWindow":16384,"status":"available","releasedDate":"2024-02-28","homepageUrl":"https://huggingface.co/blog/starcoder2","modelCardUrl":"https://huggingface.co/bigcode/starcoder2-15b","paperUrl":"https://arxiv.org/abs/2402.19173","notable":false,"org":{"id":"c151c345-7a7f-486e-8c90-b34fd5741729","slug":"bigcode","name":"BigCode","shortName":"BigCode","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"182425cc-7450-463b-914f-8efc1aab3d4b","slug":"minicpm-v-2-6","name":"MiniCPM-V 2.6","family":"MiniCPM-V","version":"2.6","description":"8B vision-language model for local image, multi-image, OCR, and video understanding, with llama.cpp and Ollama support.","orgSlug":"openbmb","isFrontier":false,"licenseType":"open_weight","licenseName":"MiniCPM Model License","weightsAvailable":true,"modalities":["text","vision","video"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-08-02","homepageUrl":"https://huggingface.co/openbmb/MiniCPM-V-2_6","modelCardUrl":"https://huggingface.co/openbmb/MiniCPM-V-2_6","paperUrl":"https://arxiv.org/abs/2408.01800","notable":true,"org":{"id":"2b61ee3c-d446-4edb-ac7d-0bcec6804223","slug":"openbmb","name":"OpenBMB","shortName":"OpenBMB","orgType":"research_org","countryCode":"CN","countryName":"China"}},{"id":"fb8ebb0d-f3b1-45c2-8b7f-9e1e502af9ba","slug":"llava-1-5-13b","name":"LLaVA 1.5 13B","family":"LLaVA","version":"1.5 13B","description":"Open vision-language assistant and one of the most widely run early local multimodal models.","orgSlug":"llava","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 2 / research license","weightsAvailable":true,"modalities":["text","vision"],"architecture":"hybrid","paramsTotalB":13,"paramsActiveB":13,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-09-30","homepageUrl":"https://llava-vl.github.io/","modelCardUrl":"https://huggingface.co/llava-hf/llava-1.5-13b-hf","paperUrl":"https://arxiv.org/abs/2310.03744","notable":false,"org":{"id":"d5d0746c-e17f-4d2e-bac8-bfe2d1c10914","slug":"llava","name":"LLaVA","shortName":"LLaVA","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"753ad65b-00cb-47d0-be31-2d0093ada16a","slug":"tinyllama-1-1b-chat","name":"TinyLlama 1.1B Chat","family":"TinyLlama","version":"1.1B Chat","description":"Compact Llama-style 1.1B chat model trained for local experimentation and low-memory deployments.","orgSlug":"tinyllama","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":1.1,"paramsActiveB":1.1,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-01-01","homepageUrl":"https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0","modelCardUrl":"https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0","notable":false,"org":{"id":"a4915eef-1285-4445-b435-1dae215b0918","slug":"tinyllama","name":"TinyLlama","shortName":"TinyLlama","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"3f6db59c-dcd6-435b-9f59-7a82b5e92809","slug":"dolphin-3-0-llama3-1-8b","name":"Dolphin 3.0 Llama 3.1 8B","family":"Dolphin","version":"3.0 Llama 3.1 8B","description":"Popular local assistant model tuned for coding, math, function calling, and agentic workflows.","orgSlug":"cognitive-computations","isFrontier":false,"licenseType":"open_weight","licenseName":"Llama 3.1 Community License","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":8,"paramsActiveB":8,"isSizeDisclosed":true,"contextWindow":128000,"status":"available","releasedDate":"2025-02-02","homepageUrl":"https://huggingface.co/dphn/Dolphin3.0-Llama3.1-8B","modelCardUrl":"https://huggingface.co/dphn/Dolphin3.0-Llama3.1-8B","notable":false,"org":{"id":"eab8da5b-a91d-4efb-86e7-7043003a286e","slug":"cognitive-computations","name":"Cognitive Computations","shortName":"Cognitive","orgType":"open_weight_lab","countryCode":"US","countryName":"United States"}},{"id":"8b788920-d1e2-4d19-af79-d5b9edd8f483","slug":"lfm2-1-2b","name":"LFM2 1.2B","family":"LFM2","version":"1.2B","description":"Liquid AI hybrid model for efficient CPU/GPU/NPU local deployment, using short convolutions plus attention blocks.","orgSlug":"liquid-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"LFM Open License v1.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"hybrid","paramsTotalB":1.17,"paramsActiveB":1.17,"isSizeDisclosed":true,"contextWindow":32768,"status":"available","releasedDate":"2025-11-28","homepageUrl":"https://huggingface.co/LiquidAI/LFM2-1.2B","modelCardUrl":"https://huggingface.co/LiquidAI/LFM2-1.2B","paperUrl":"https://arxiv.org/abs/2511.23404","notable":false,"org":{"id":"b98d12b3-586f-443c-863a-10ae93f854f4","slug":"liquid-ai","name":"Liquid AI","shortName":"Liquid","orgType":"startup","countryCode":"US","countryName":"United States"}},{"id":"daf72799-c09a-4816-aa22-d303e3c5a402","slug":"dbrx-instruct","name":"DBRX Instruct","family":"DBRX","version":"Instruct","description":"Databricks' 132B-total / 36B-active open MoE model for code, math, RAG, and enterprise self-hosted workloads.","orgSlug":"databricks","isFrontier":false,"licenseType":"open_weight","licenseName":"Databricks Open Model License","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":132,"paramsActiveB":36,"isSizeDisclosed":true,"contextWindow":32000,"status":"available","releasedDate":"2024-03-27","homepageUrl":"https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm","modelCardUrl":"https://huggingface.co/databricks/dbrx-instruct","notable":true,"org":{"id":"9fba6ae8-74a1-424d-a427-ee185e8fcd46","slug":"databricks","name":"Databricks / MosaicML","shortName":"Databricks","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"fe018ddb-23ba-4200-a044-ec366671c254","slug":"mpt-7b","name":"MPT-7B","family":"MPT","version":"7B","description":"MosaicML's permissively licensed 7B model, an early favorite for commercial local fine-tuning and long-context variants.","orgSlug":"databricks","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"contextWindow":2048,"status":"available","releasedDate":"2023-05-05","homepageUrl":"https://www.databricks.com/blog/mpt-7b","modelCardUrl":"https://huggingface.co/mosaicml/mpt-7b-instruct","notable":false,"org":{"id":"9fba6ae8-74a1-424d-a427-ee185e8fcd46","slug":"databricks","name":"Databricks / MosaicML","shortName":"Databricks","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"26ba038e-c49b-428d-9d5a-ccc584aa858b","slug":"snowflake-arctic","name":"Snowflake Arctic","family":"Arctic","version":"480B","description":"Apache-2.0 enterprise LLM with 480B total / 17B active parameters, optimized for SQL, code, and instruction following.","orgSlug":"snowflake","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"moe","paramsTotalB":480,"paramsActiveB":17,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-04-24","homepageUrl":"https://www.snowflake.com/en/blog/arctic-open-efficient-foundation-language-models-snowflake/","modelCardUrl":"https://huggingface.co/Snowflake/snowflake-arctic-instruct","notable":false,"org":{"id":"cc3fab15-bf33-4045-9c81-e726ec53904f","slug":"snowflake","name":"Snowflake AI Research","shortName":"Snowflake","orgType":"big_tech","countryCode":"US","countryName":"United States"}},{"id":"a410ab4a-adbc-4d39-acdd-7473cf1cebc9","slug":"stablelm-2-1-6b","name":"Stable LM 2 1.6B","family":"Stable LM","version":"2 1.6B","description":"Small multilingual Stable LM release built for low hardware barriers and local experimentation.","orgSlug":"stability-ai","isFrontier":false,"licenseType":"open_weight","licenseName":"Stability AI Membership","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":1.6,"paramsActiveB":1.6,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-01-19","homepageUrl":"https://stability.ai/news-updates/introducing-stable-lm-2","modelCardUrl":"https://huggingface.co/stabilityai/stablelm-2-1_6b-chat","paperUrl":"https://arxiv.org/abs/2402.17834","notable":false,"org":{"id":"805334a3-85b4-4c2d-8f7c-49d2daed864f","slug":"stability-ai","name":"Stability AI","shortName":"Stability","orgType":"startup","countryCode":"GB","countryName":"United Kingdom"}},{"id":"ebfe1971-4c68-428b-984e-22df2506f1ee","slug":"vicuna-13b","name":"Vicuna 13B","family":"Vicuna","version":"13B","description":"LMSYS instruction-tuned LLaMA model that became a landmark early local ChatGPT-style assistant.","orgSlug":"lmsys","isFrontier":false,"licenseType":"open_weight","licenseName":"LLaMA Research License","weightsAvailable":true,"modalities":["text"],"architecture":"dense","paramsTotalB":13,"paramsActiveB":13,"isSizeDisclosed":true,"status":"available","releasedDate":"2023-03-30","homepageUrl":"https://lmsys.org/blog/2023-03-30-vicuna/","modelCardUrl":"https://huggingface.co/lmsys/vicuna-13b-v1.3","notable":false,"org":{"id":"5ef26afe-0558-40d2-9e23-5936188f47d1","slug":"lmsys","name":"LMSYS / SkyLab","shortName":"LMSYS","orgType":"research_org","countryCode":"US","countryName":"United States"}},{"id":"bac94f85-074f-4d10-9bc3-d5694d60a44d","slug":"openchat-3-5","name":"OpenChat 3.5","family":"OpenChat","version":"3.5","description":"Compact Mistral-based local chat model trained with C-RLFT, popular in early 2024 local leaderboards.","orgSlug":"openchat","isFrontier":false,"licenseType":"open_source","licenseName":"Apache-2.0","weightsAvailable":true,"modalities":["text","code"],"architecture":"dense","paramsTotalB":7,"paramsActiveB":7,"isSizeDisclosed":true,"status":"available","releasedDate":"2024-01-06","homepageUrl":"https://huggingface.co/openchat/openchat-3.5-0106","modelCardUrl":"https://huggingface.co/openchat/openchat-3.5-0106","notable":false,"org":{"id":"a542506c-07d6-4922-8d08-8a01a8ad628d","slug":"openchat","name":"OpenChat","shortName":"OpenChat","orgType":"open_weight_lab","countryCode":"US","countryName":"United States"}}],"events":[{"id":"3d3b126b-6e7c-47c1-b3bd-8818e0f10931","modelSlug":"kimi-k2-7-code","eventType":"released","eventDate":"2026-06-18","title":"Moonshot releases Kimi K2.7 Code","summary":"Open coding-focused Kimi model with 1T total / 32B active parameters, native image/video input, and always-on thinking mode.","sourceUrl":"https://www.kimi.com/resources/kimi-k2-7-code","modelName":"Kimi K2.7 Code","orgName":"Moonshot AI","countryCode":"CN"},{"id":"123bca4c-dfc1-4607-b078-838c4e34bc3f","modelSlug":"glm-5-2","eventType":"released","eventDate":"2026-06-17","title":"Z.ai releases GLM-5.2 with a 1M-token context","summary":"MIT-licensed GLM flagship focused on long-horizon coding, agentic engineering, and IndexShare sparse-attention reuse.","sourceUrl":"https://huggingface.co/zai-org/GLM-5.2","modelName":"GLM-5.2","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"aa643a79-8601-4d0c-b8e2-af1a1528c16b","modelSlug":"minimax-m3","eventType":"released","eventDate":"2026-06-16","title":"MiniMax releases MiniMax-M3","summary":"Native multimodal 428B/23B-active model with one-million-token context and MiniMax Sparse Attention.","sourceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M3","modelName":"MiniMax-M3","orgName":"MiniMax","countryCode":"CN"},{"id":"fc003c90-cf29-4ad7-a762-0fed70fa5a09","modelSlug":"claude-fable-5","eventType":"withdrawn","eventDate":"2026-06-12","title":"US government orders Anthropic to pull Fable 5 and Mythos 5","summary":"Access suspended three days after launch under an export-control directive citing national security.","sourceUrl":"https://thenewstack.io/us-gov-orders-anthropic-to-pull-fable-5-and-mythos-5-three-days-after-launch/","modelName":"Claude Fable 5","orgName":"Anthropic","countryCode":"US"},{"id":"a9fa5466-d5a2-4c8c-940e-b6c4e10e8fa0","modelSlug":"claude-fable-5","eventType":"released","eventDate":"2026-06-09","title":"Claude Fable 5 released — first public Mythos-class model","summary":"Available across the Claude API, AWS, and Microsoft Foundry.","sourceUrl":"https://fortune.com/2026/06/09/anthropic-releases-its-first-mythos-model-to-the-public/","modelName":"Claude Fable 5","orgName":"Anthropic","countryCode":"US"},{"id":"94571211-348c-4b5c-a2cc-f82b065aa10b","modelSlug":"gpt-5-6","eventType":"announced","eventDate":"2026-06-09","title":"GPT-5.6 announced with a 1.5M-token context window","summary":"OpenAI previewed GPT-5.6, claiming the largest context window of any frontier model.","sourceUrl":"https://openai.com","modelName":"GPT-5.6","orgName":"OpenAI","countryCode":"US"},{"id":"5f265b86-4093-474d-bdfe-cefac1548785","modelSlug":"nemotron-3-ultra-550b-a55b","eventType":"released","eventDate":"2026-06-04","title":"NVIDIA releases Nemotron 3 Ultra 550B-A55B","summary":"Largest Nemotron 3 model appears on NVIDIA NIM with downloadable weights, 1M context, and agentic reasoning positioning.","sourceUrl":"https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b","modelName":"Nemotron 3 Ultra 550B-A55B","orgName":"NVIDIA","countryCode":"US"},{"id":"8bba20a0-c839-4786-b656-cfade034d7b6","modelSlug":"claude-opus-4-8","eventType":"released","eventDate":"2026-05-28","title":"Claude Opus 4.8 released","summary":"Agentic upgrades and stronger long-running task performance.","sourceUrl":"https://anthropic.com","modelName":"Claude Opus 4.8","orgName":"Anthropic","countryCode":"US"},{"id":"ba6dc896-66d1-4a94-8459-0ee11a9febf4","modelSlug":"minimax-m2-7","eventType":"released","eventDate":"2026-05-26","title":"MiniMax releases MiniMax-M2.7","summary":"Open-weight agentic model focused on software engineering, productivity tasks, and model self-evolution workflows.","sourceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M2.7","modelName":"MiniMax-M2.7","orgName":"MiniMax","countryCode":"CN"},{"id":"4c9060e7-bd48-4e7f-affc-4a7e298b7d84","modelSlug":"gemini-3-5-pro","eventType":"announced","eventDate":"2026-05-19","title":"Gemini 3.5 Pro announced at Google I/O 2026","sourceUrl":"https://deepmind.google","modelName":"Gemini 3.5 Pro","orgName":"Google DeepMind","countryCode":"US"},{"id":"6937c6e2-bdf4-4192-a46b-c56410d1f72c","modelSlug":"qwen3-6-27b","eventType":"released","eventDate":"2026-05-12","title":"Qwen releases Qwen3.6-27B","sourceUrl":"https://qwen.ai","modelName":"Qwen3.6-27B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"e0d88127-6a8f-45f5-b644-b922c784fee0","modelSlug":"claude-3-7-sonnet","eventType":"retired","eventDate":"2026-05-11","title":"Claude 3.7 Sonnet retired","summary":"Endpoint shut down as part of Anthropic's 2026 deprecation calendar.","sourceUrl":"https://anthropic.com","modelName":"Claude 3.7 Sonnet","orgName":"Anthropic","countryCode":"US"},{"id":"0146d22c-3b90-45b1-be82-671b77f0c215","modelSlug":"grok-4-3","eventType":"released","eventDate":"2026-05-06","title":"Grok 4.3 generally available","summary":"1M-token context at $1.25/$2.50 per million tokens; also live on Microsoft Foundry.","sourceUrl":"https://x.ai","modelName":"Grok 4.3","orgName":"xAI","countryCode":"US"},{"id":"8ac60e88-64c1-4547-8b7c-82ed299fa83d","modelSlug":"deepseek-v4-flash","eventType":"released","eventDate":"2026-04-24","title":"DeepSeek V4-Flash released as the efficient V4 variant","sourceUrl":"https://api-docs.deepseek.com/news/news260424","modelName":"DeepSeek V4-Flash","orgName":"DeepSeek","countryCode":"CN"},{"id":"7207a2fc-636e-423c-b8dc-c3df246418ef","modelSlug":"deepseek-v4-pro","eventType":"released","eventDate":"2026-04-24","title":"DeepSeek V4-Pro preview released with 1M context","summary":"DeepSeek introduced the V4 preview series under MIT, led by V4-Pro at 1.6T total / 49B active parameters.","sourceUrl":"https://api-docs.deepseek.com/news/news260424","modelName":"DeepSeek V4-Pro","orgName":"DeepSeek","countryCode":"CN"},{"id":"e708a840-f7a4-4936-acb8-25216f108b71","modelSlug":"hunyuan-a13b-instruct","eventType":"released","eventDate":"2026-04-22","title":"Tencent releases Hunyuan-A13B-Instruct","summary":"80B/13B-active Hunyuan MoE model released with open weights and agentic tool-use support.","sourceUrl":"https://huggingface.co/tencent/Hunyuan-A13B-Instruct","modelName":"Hunyuan-A13B-Instruct","orgName":"Tencent Hunyuan","countryCode":"CN"},{"id":"3d28ec55-d93f-4143-8e33-1a89f331738f","modelSlug":"glm-5-1","eventType":"released","eventDate":"2026-04-08","title":"Z.ai releases GLM-5.1 for long-horizon agentic engineering","sourceUrl":"https://z.ai/blog/glm-5.1","modelName":"GLM-5.1","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"6d4f8eeb-65e7-4cac-9afa-2c6fdffec5c6","modelSlug":"claude-mythos","eventType":"announced","eventDate":"2026-04-07","title":"Anthropic discloses Claude Mythos but withholds public release","summary":"Frontier model shipped only to ~50 defensive-security partners via Project Glasswing.","sourceUrl":"https://www.axios.com/2026/06/09/anthropic-mythos-claude-fable-5","modelName":"Claude Mythos","orgName":"Anthropic","countryCode":"US"},{"id":"a2ec3ac0-fce7-417c-8c3f-da42f0cb27c5","modelSlug":"gpt-4o","eventType":"retired","eventDate":"2026-04-03","title":"GPT-4o fully retired from ChatGPT","summary":"Removed from all ChatGPT plans after a Feb 13 deprecation notice.","sourceUrl":"https://help.openai.com/en/articles/20001051-retiring-gpt-4o-and-other-chatgpt-models","modelName":"GPT-4o","orgName":"OpenAI","countryCode":"US"},{"id":"65286edd-a1dd-46c3-aeed-0509776f9a3a","modelSlug":"gemma-4-31b","eventType":"released","eventDate":"2026-04-02","title":"Google DeepMind releases Gemma 4","summary":"Gemma 4 introduces advanced reasoning open models in 12B, 26B, and 31B sizes.","sourceUrl":"https://deepmind.google/models/gemma/","modelName":"Gemma 4 31B","orgName":"Google DeepMind","countryCode":"US"},{"id":"85a1e271-3213-4b5a-a16d-f2132b565c69","modelSlug":"kimi-k2-6","eventType":"released","eventDate":"2026-03-30","title":"Kimi K2.6 released","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2.6","modelName":"Kimi K2.6","orgName":"Moonshot AI","countryCode":"CN"},{"id":"b1141e00-fcbd-4834-b200-fdad610f568f","modelSlug":"mistral-medium-3-5","eventType":"released","eventDate":"2026-03-18","title":"Mistral Medium 3.5 released","sourceUrl":"https://mistral.ai","modelName":"Mistral Medium 3.5","orgName":"Mistral AI","countryCode":"FR"},{"id":"15614fbd-be7d-46eb-a487-e9cde9786b17","modelSlug":"nemotron-3-super-120b-a12b","eventType":"released","eventDate":"2026-03-16","title":"NVIDIA releases Nemotron 3 Super 120B-A12B","sourceUrl":"https://build.nvidia.com/nvidia/nemotron-3-super-120b-a12b","modelName":"Nemotron 3 Super 120B-A12B","orgName":"NVIDIA","countryCode":"US"},{"id":"e19276d0-cbf3-412f-84d4-8345f142c1ad","modelSlug":"step-3-5-flash","eventType":"released","eventDate":"2026-03-14","title":"StepFun releases Step-3.5-Flash under Apache-2.0","sourceUrl":"https://huggingface.co/stepfun-ai/Step-3.5-Flash","modelName":"Step-3.5-Flash","orgName":"StepFun","countryCode":"CN"},{"id":"f41e0208-b4b1-4558-9d06-463aabd35b97","modelSlug":"sarvam-105b","eventType":"released","eventDate":"2026-03-06","title":"Sarvam AI open-sources Sarvam-105B","summary":"Apache-2.0 MoE model focused on reasoning, coding, agentic tasks, and Indian-language performance.","sourceUrl":"https://www.sarvam.ai/blogs/sarvam-30b-105b","modelName":"Sarvam-105B","orgName":"Sarvam AI","countryCode":"IN"},{"id":"05bb40ae-562a-4f57-a079-aa000986668b","modelSlug":"gpt-5-4","eventType":"released","eventDate":"2026-03-05","title":"GPT-5.4 and GPT-5.4 Thinking released","sourceUrl":"https://openai.com","modelName":"GPT-5.4","orgName":"OpenAI","countryCode":"US"},{"id":"05a286c0-d409-4d0c-b490-eef33783b220","modelSlug":"qwen3-5","eventType":"released","eventDate":"2026-02-20","title":"Qwen3.5-397B released (Apache-2.0, 201 languages)","sourceUrl":"https://qwen.ai","modelName":"Qwen3.5-397B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"62a56dc4-e674-4392-bb64-29b2e133afc4","modelSlug":"gemini-3-1-pro","eventType":"released","eventDate":"2026-02-19","title":"Gemini 3.1 Pro generally available","sourceUrl":"https://deepmind.google","modelName":"Gemini 3.1 Pro","orgName":"Google DeepMind","countryCode":"US"},{"id":"72fbf142-7ef4-4b40-be47-728ddff0b04a","modelSlug":"glm-5","eventType":"released","eventDate":"2026-02-11","title":"Z.ai releases GLM-5 for complex systems engineering","sourceUrl":"https://z.ai/blog/glm-5","modelName":"GLM-5","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"39f6d104-252d-4e4b-9591-bcbccbeafc8f","modelSlug":"claude-opus-4-6","eventType":"released","eventDate":"2026-02-05","title":"Claude Opus 4.6 released with autonomous coding","sourceUrl":"https://anthropic.com","modelName":"Claude Opus 4.6","orgName":"Anthropic","countryCode":"US"},{"id":"6a2dafd7-3eca-49bc-9adf-051ea8db00e3","modelSlug":"qwen3-coder-next","eventType":"released","eventDate":"2026-02-03","title":"Qwen releases Qwen3-Coder-Next for coding agents","sourceUrl":"https://huggingface.co/Qwen/Qwen3-Coder-Next","modelName":"Qwen3-Coder-Next","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"b3904959-38df-4944-8aca-e1acd9ac7d78","modelSlug":"kimi-k2-5","eventType":"released","eventDate":"2026-01-27","title":"Moonshot releases Kimi K2.5","summary":"Open multimodal K2 upgrade with MoonViT, thinking modes, visual coding, and agent-swarm workflows.","sourceUrl":"https://www.kimi.com/ai-models/kimi-k2-5","modelName":"Kimi K2.5","orgName":"Moonshot AI","countryCode":"CN"},{"id":"10fd142f-e931-4530-a079-8adb1497cdfa","modelSlug":"glm-4-7","eventType":"released","eventDate":"2026-01-08","title":"Z.ai releases GLM-4.7 with stronger coding and tool use","sourceUrl":"https://z.ai/blog/glm-4.7","modelName":"GLM-4.7","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"2e35d89f-a2b3-4a3f-a391-672839b51e4c","modelSlug":"olmo-3-think-32b","eventType":"released","eventDate":"2025-12-15","title":"Ai2 releases OLMo 3 Think 32B as a fully open reasoning model","sourceUrl":"https://arxiv.org/abs/2512.13961","modelName":"OLMo 3 Think 32B","orgName":"Allen Institute for AI (Ai2)","countryCode":"US"},{"id":"e3c608bd-75a7-450b-bb15-0946c0a0f50b","modelSlug":"nemotron-3-nano-30b-a3b","eventType":"released","eventDate":"2025-12-15","title":"NVIDIA launches the Nemotron 3 open model family","sourceUrl":"https://build.nvidia.com/nvidia/nemotron-3-nano-30b-a3b","modelName":"Nemotron 3 Nano 30B-A3B","orgName":"NVIDIA","countryCode":"US"},{"id":"daeedeec-62e7-4f02-aac9-4955e876a0f6","modelSlug":"glm-4-6v","eventType":"released","eventDate":"2025-12-08","title":"Z.ai releases GLM-4.6V with native multimodal function calling","sourceUrl":"https://z.ai/blog/glm-4.6v","modelName":"GLM-4.6V","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"b2e9504a-516c-4f6c-95dc-afae13cf2dd7","modelSlug":"mistral-large-3","eventType":"released","eventDate":"2025-12-02","title":"Mistral 3 family released (open-weight Large 3 + small dense models)","sourceUrl":"https://mistral.ai/news/mistral-3/","modelName":"Mistral Large 3","orgName":"Mistral AI","countryCode":"FR"},{"id":"6dafe073-6886-4f7e-8bd9-5322d67bdf32","modelSlug":"deepseek-v3-2-speciale","eventType":"released","eventDate":"2025-12-01","title":"DeepSeek-V3.2-Speciale released for high-compute reasoning","sourceUrl":"https://api-docs.deepseek.com/news/news251201","modelName":"DeepSeek-V3.2-Speciale","orgName":"DeepSeek","countryCode":"CN"},{"id":"4cacee49-5ec7-4eff-83d3-e94d620ea9ce","modelSlug":"deepseek-v3-2","eventType":"released","eventDate":"2025-12-01","title":"DeepSeek-V3.2 released for reasoning-first agent workflows","sourceUrl":"https://api-docs.deepseek.com/news/news251201","modelName":"DeepSeek-V3.2","orgName":"DeepSeek","countryCode":"CN"},{"id":"f7518326-ce42-41a3-8fb7-8b35d80cd345","modelSlug":"lfm2-1-2b","eventType":"released","eventDate":"2025-11-28","title":"Liquid AI releases LFM2 1.2B","sourceUrl":"https://huggingface.co/LiquidAI/LFM2-1.2B","modelName":"LFM2 1.2B","orgName":"Liquid AI","countryCode":"US"},{"id":"171d646e-e29e-44b0-a990-109b68baa41f","modelSlug":"kimi-k2-thinking","eventType":"released","eventDate":"2025-11-06","title":"Moonshot releases Kimi K2 Thinking","summary":"Open K2 reasoning-agent variant for deep thinking and stable long-horizon tool orchestration.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Thinking","modelName":"Kimi K2 Thinking","orgName":"Moonshot AI","countryCode":"CN"},{"id":"668ef9f2-f2db-4bfc-866c-3ff7f7128f3b","modelSlug":"kimi-linear-48b-a3b-instruct","eventType":"released","eventDate":"2025-10-31","title":"Moonshot releases Kimi Linear 48B-A3B","summary":"MIT-licensed hybrid linear-attention checkpoints with a 1M-token context and lower KV-cache use.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-Linear-48B-A3B-Instruct","modelName":"Kimi-Linear-48B-A3B-Instruct","orgName":"Moonshot AI","countryCode":"CN"},{"id":"f8eba520-d433-4a30-ab08-2794b8e54ae3","modelSlug":"glm-4-6","eventType":"released","eventDate":"2025-09-30","title":"Z.ai releases GLM-4.6 with a 200K-token context window","sourceUrl":"https://z.ai/blog/glm-4.6","modelName":"GLM-4.6","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"1fa83ba4-ba03-4773-973d-7d9c28b4f379","modelSlug":"deepseek-v3-2-exp","eventType":"released","eventDate":"2025-09-29","title":"DeepSeek-V3.2-Exp released to preview sparse attention","sourceUrl":"https://api-docs.deepseek.com/news/news250929","modelName":"DeepSeek-V3.2-Exp","orgName":"DeepSeek","countryCode":"CN"},{"id":"65757089-18ef-40c3-9e69-4b7a0cc15c5f","modelSlug":"deepseek-v3-1-terminus","eventType":"updated","eventDate":"2025-09-22","title":"DeepSeek-V3.1-Terminus updates V3.1 stability and agents","sourceUrl":"https://api-docs.deepseek.com/news/news250922","modelName":"DeepSeek-V3.1-Terminus","orgName":"DeepSeek","countryCode":"CN"},{"id":"3282a206-beb5-4714-8644-659e1dc8f519","modelSlug":"kimi-k2-instruct-0905","eventType":"released","eventDate":"2025-09-05","title":"Moonshot updates Kimi K2 Instruct with 256K context","summary":"The 0905 update improves agentic coding and frontend generation while doubling context length.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct-0905","modelName":"Kimi K2 Instruct 0905","orgName":"Moonshot AI","countryCode":"CN"},{"id":"6f1e7166-5c32-4095-99bd-4ef0c991b3d6","modelSlug":"gemma-3-27b","eventType":"released","eventDate":"2025-09-04","title":"Google releases Gemma 3 (multimodal, 128k context)","sourceUrl":"https://developers.googleblog.com/en/introducing-gemma3/","modelName":"Gemma 3 27B","orgName":"Google DeepMind","countryCode":"US"},{"id":"15051787-eb7d-4c33-83fd-b683eb801e9d","modelSlug":"deepseek-v3-1","eventType":"released","eventDate":"2025-08-21","title":"DeepSeek-V3.1 released with hybrid thinking modes","sourceUrl":"https://api-docs.deepseek.com/news/news250821","modelName":"DeepSeek-V3.1","orgName":"DeepSeek","countryCode":"CN"},{"id":"8358b399-33d4-465c-a871-22e4663b5438","modelSlug":"seed-oss-36b-instruct","eventType":"released","eventDate":"2025-08-20","title":"ByteDance Seed releases Seed-OSS-36B-Instruct under Apache-2.0","sourceUrl":"https://huggingface.co/ByteDance-Seed/Seed-OSS-36B-Instruct","modelName":"Seed-OSS-36B-Instruct","orgName":"ByteDance Seed","countryCode":"CN"},{"id":"8703f9b4-6e61-4b89-baac-46e749b5f434","modelSlug":"glm-4-5v","eventType":"released","eventDate":"2025-08-11","title":"Z.ai releases GLM-4.5V for multimodal reasoning","sourceUrl":"https://z.ai/blog/glm-4.5v","modelName":"GLM-4.5V","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"cdbc0364-f879-418f-bd52-36856b9fcb96","modelSlug":"gpt-oss-20b","eventType":"released","eventDate":"2025-08-05","title":"OpenAI releases gpt-oss-20b","sourceUrl":"https://openai.com/index/introducing-gpt-oss/","modelName":"gpt-oss-20b","orgName":"OpenAI","countryCode":"US"},{"id":"1504862a-e8ad-4c73-84ad-ce4f62f37c99","modelSlug":"gpt-oss-120b","eventType":"released","eventDate":"2025-08-05","title":"OpenAI releases gpt-oss-120b","sourceUrl":"https://openai.com/index/introducing-gpt-oss/","modelName":"gpt-oss-120b","orgName":"OpenAI","countryCode":"US"},{"id":"6b42481f-3e00-4860-ae78-38bc8048b40f","modelSlug":"falcon-h1-34b","eventType":"released","eventDate":"2025-07-31","title":"TII releases Falcon-H1 hybrid attention-SSM models","sourceUrl":"https://falcon-lm.github.io/blog/falcon-h1/","modelName":"Falcon-H1 34B","orgName":"Technology Innovation Institute","countryCode":"AE"},{"id":"a3ac0117-cc55-4c0e-b22e-fbf8675ab829","modelSlug":"glm-4-5","eventType":"released","eventDate":"2025-07-28","title":"Z.ai releases GLM-4.5 under MIT","sourceUrl":"https://z.ai/blog/glm-4.5","modelName":"GLM-4.5","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"72575810-0f5f-472a-b4c3-aa195c787f8b","modelSlug":"glm-4-5-air","eventType":"released","eventDate":"2025-07-28","title":"Z.ai releases GLM-4.5-Air as the efficient GLM-4.5 variant","sourceUrl":"https://z.ai/blog/glm-4.5","modelName":"GLM-4.5-Air","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"fae8e0a9-da2c-4188-9122-0ef894746b8d","modelSlug":"exaone-4-0-32b","eventType":"released","eventDate":"2025-07-15","title":"LG AI Research releases EXAONE 4.0 32B","sourceUrl":"https://arxiv.org/abs/2507.11407","modelName":"EXAONE 4.0 32B","orgName":"LG AI Research","countryCode":"KR"},{"id":"66ed7c0d-e29d-431d-be22-52196f01be52","modelSlug":"kimi-k2-instruct","eventType":"released","eventDate":"2025-07-11","title":"Moonshot releases Kimi K2 Instruct","summary":"Original open 1T-parameter K2 MoE release optimized for coding, reasoning, and agentic tool use.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-K2-Instruct","modelName":"Kimi K2 Instruct","orgName":"Moonshot AI","countryCode":"CN"},{"id":"7394717e-c18e-4403-8182-d3312a893133","modelSlug":"grok-4","eventType":"released","eventDate":"2025-07-09","title":"xAI releases Grok 4","sourceUrl":"https://x.ai/news/grok-4","modelName":"Grok 4","orgName":"xAI","countryCode":"US"},{"id":"41e8a3eb-aa35-406f-9bb2-80728de89f4d","modelSlug":"smollm3-3b","eventType":"released","eventDate":"2025-07-08","title":"Hugging Face releases SmolLM3 3B","sourceUrl":"https://huggingface.co/blog/smollm3","modelName":"SmolLM3 3B","orgName":"Hugging Face","countryCode":"US"},{"id":"bf00567a-50a3-4ae1-b86c-c4e9eb762396","modelSlug":"ernie-4-5-300b-a47b","eventType":"released","eventDate":"2025-06-30","title":"Baidu open-sources ERNIE-4.5 language models","sourceUrl":"https://github.com/PaddlePaddle/ERNIE","modelName":"ERNIE-4.5-300B-A47B","orgName":"Baidu","countryCode":"CN"},{"id":"f371d73d-8ecc-4312-93e4-15474c640ecd","modelSlug":"ernie-4-5-vl-424b-a47b","eventType":"released","eventDate":"2025-06-30","title":"Baidu open-sources ERNIE-4.5 vision-language models","sourceUrl":"https://github.com/PaddlePaddle/ERNIE","modelName":"ERNIE-4.5-VL-424B-A47B","orgName":"Baidu","countryCode":"CN"},{"id":"d314b437-561e-4a2c-9db2-93d60df09b1b","modelSlug":"kimi-vl-a3b-thinking-2506","eventType":"released","eventDate":"2025-06-21","title":"Moonshot releases Kimi-VL-A3B-Thinking-2506","summary":"Updated efficient multimodal reasoning model with stronger video, high-resolution perception, and lower thinking-token use.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-VL-A3B-Thinking-2506","modelName":"Kimi-VL-A3B-Thinking-2506","orgName":"Moonshot AI","countryCode":"CN"},{"id":"56b3131b-6be8-4013-8585-3d796f8414e8","modelSlug":"kimi-dev-72b","eventType":"released","eventDate":"2025-06-17","title":"Moonshot releases Kimi-Dev-72B","summary":"Open coding LLM trained with repository-level reinforcement learning for issue resolution.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-Dev-72B","modelName":"Kimi-Dev-72B","orgName":"Moonshot AI","countryCode":"CN"},{"id":"f56b62d8-483f-46c2-a148-dedb6e020a86","modelSlug":"minimax-m1-80k","eventType":"released","eventDate":"2025-06-16","title":"MiniMax releases MiniMax-M1-80k with 1M-token context","sourceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-M1-80k","modelName":"MiniMax-M1-80k","orgName":"MiniMax","countryCode":"CN"},{"id":"1e8e055b-f05d-4df7-943b-81ca9adfdd74","modelSlug":"magistral-small","eventType":"released","eventDate":"2025-06-10","title":"Mistral releases Magistral Small","sourceUrl":"https://mistral.ai/news/magistral/","modelName":"Magistral Small","orgName":"Mistral AI","countryCode":"FR"},{"id":"b039ef60-1e21-468b-9e1c-6c82d058417d","modelSlug":"magistral-medium","eventType":"released","eventDate":"2025-06-10","title":"Mistral releases Magistral Medium","sourceUrl":"https://mistral.ai/news/magistral/","modelName":"Magistral Medium","orgName":"Mistral AI","countryCode":"FR"},{"id":"49b80648-b2ea-4828-85cc-f89368fa2e53","modelSlug":"deepseek-r1-0528","eventType":"released","eventDate":"2025-05-28","title":"DeepSeek-R1-0528 released with stronger reasoning and function calling","sourceUrl":"https://api-docs.deepseek.com/news/news250528","modelName":"DeepSeek-R1-0528","orgName":"DeepSeek","countryCode":"CN"},{"id":"93afb3e0-841f-4e3b-8ffa-2c196f09ed5a","modelSlug":"seed-thinking-v1-5","eventType":"released","eventDate":"2025-05-22","title":"ByteDance Seed releases Seed Thinking v1.5","sourceUrl":"https://seed.bytedance.com/","modelName":"Seed Thinking v1.5","orgName":"ByteDance Seed","countryCode":"CN"},{"id":"3586721b-8cc3-4399-ba50-871b41c037aa","modelSlug":"claude-opus-4","eventType":"released","eventDate":"2025-05-22","title":"Anthropic releases Claude Opus 4","sourceUrl":"https://www.anthropic.com/news/claude-4","modelName":"Claude Opus 4","orgName":"Anthropic","countryCode":"US"},{"id":"35bd7e2d-6b99-400b-b270-97ed19809995","modelSlug":"sarvam-m","eventType":"released","eventDate":"2025-05-21","title":"Sarvam AI releases Sarvam-M","sourceUrl":"https://www.sarvam.ai/blogs/sarvam-m","modelName":"Sarvam-M","orgName":"Sarvam AI","countryCode":"IN"},{"id":"68eb89ca-970f-44f8-bb24-bc6ba72eb498","modelSlug":"granite-3-3-8b","eventType":"released","eventDate":"2025-04-30","title":"IBM releases Granite 3.3 8B","sourceUrl":"https://huggingface.co/ibm-granite/granite-3.3-8b-instruct","modelName":"Granite 3.3 8B","orgName":"IBM","countryCode":"US"},{"id":"8bfcb760-6bd2-4971-9a31-f76e0f4eabb4","modelSlug":"phi-4-reasoning","eventType":"released","eventDate":"2025-04-30","title":"Microsoft releases Phi-4 Reasoning","sourceUrl":"https://huggingface.co/microsoft/Phi-4-reasoning","modelName":"Phi-4 Reasoning","orgName":"Microsoft","countryCode":"US"},{"id":"5e75ff71-a8a8-4ab9-a2d8-587603f4e4cb","modelSlug":"qwen3-235b-a22b","eventType":"released","eventDate":"2025-04-28","title":"Qwen releases Qwen3-235B-A22B","sourceUrl":"https://qwenlm.github.io/blog/qwen3/","modelName":"Qwen3-235B-A22B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"a77c6f52-847b-4b1a-ac62-52f2547d978a","modelSlug":"kimi-audio-7b-instruct","eventType":"released","eventDate":"2025-04-25","title":"Moonshot releases Kimi-Audio-7B-Instruct","summary":"Open audio foundation model for speech recognition, audio QA, captioning, generation, and conversation.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-Audio-7B-Instruct","modelName":"Kimi-Audio-7B-Instruct","orgName":"Moonshot AI","countryCode":"CN"},{"id":"0e74057b-b67c-4eb0-8b74-aab83ee82aee","modelSlug":"kimi-vl-a3b-instruct","eventType":"released","eventDate":"2025-04-17","title":"Moonshot releases Kimi-VL-A3B-Instruct","summary":"Efficient MIT-licensed vision-language MoE for OCR, video, long documents, and agent tasks.","sourceUrl":"https://huggingface.co/moonshotai/Kimi-VL-A3B-Instruct","modelName":"Kimi-VL-A3B-Instruct","orgName":"Moonshot AI","countryCode":"CN"},{"id":"2669c40b-12dd-4a12-9824-f24e9a194147","modelSlug":"o3","eventType":"released","eventDate":"2025-04-16","title":"OpenAI releases OpenAI o3","sourceUrl":"https://openai.com/index/introducing-o3-and-o4-mini/","modelName":"OpenAI o3","orgName":"OpenAI","countryCode":"US"},{"id":"4a57f210-1ec7-4bc2-a40b-40ecf3e52bd7","modelSlug":"gpt-4-1","eventType":"released","eventDate":"2025-04-14","title":"OpenAI releases GPT-4.1","sourceUrl":"https://openai.com/index/gpt-4-1/","modelName":"GPT-4.1","orgName":"OpenAI","countryCode":"US"},{"id":"bf47b45d-32b6-4489-9b1b-e9259f71d3c1","modelSlug":"llama-4-maverick","eventType":"released","eventDate":"2025-04-05","title":"Meta releases Llama 4 Maverick","sourceUrl":"https://ai.meta.com","modelName":"Llama 4 Maverick","orgName":"Meta AI","countryCode":"US"},{"id":"cc11a0d0-bbf1-4970-a23e-daff2c6ba4a9","modelSlug":"llama-4-scout","eventType":"released","eventDate":"2025-04-05","title":"Meta releases Llama 4 Scout","sourceUrl":"https://ai.meta.com","modelName":"Llama 4 Scout","orgName":"Meta AI","countryCode":"US"},{"id":"3801d8de-7a49-49a6-aebf-c96f3a21d470","modelSlug":"llama-3-3-nemotron-super-49b","eventType":"released","eventDate":"2025-04-02","title":"NVIDIA releases Llama-3.3-Nemotron-Super-49B","sourceUrl":"https://huggingface.co/nvidia/Llama-3_3-Nemotron-Super-49B-v1","modelName":"Llama-3.3-Nemotron-Super-49B","orgName":"NVIDIA","countryCode":"US"},{"id":"9ffa0673-8901-4c60-92ad-915ea5430d9b","modelSlug":"qwen2-5-omni-7b","eventType":"released","eventDate":"2025-03-26","title":"Qwen releases Qwen2.5-Omni-7B","sourceUrl":"https://qwenlm.github.io/blog/qwen2.5-omni/","modelName":"Qwen2.5-Omni-7B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"be1a409e-dcb3-4410-a7ce-8f4835e7bf69","modelSlug":"gemini-2-5-pro","eventType":"released","eventDate":"2025-03-25","title":"Google DeepMind releases Gemini 2.5 Pro","sourceUrl":"https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/","modelName":"Gemini 2.5 Pro","orgName":"Google DeepMind","countryCode":"US"},{"id":"9452bf62-2ec6-4f69-9aa3-22db7b5e77c6","modelSlug":"deepseek-v3-0324","eventType":"released","eventDate":"2025-03-25","title":"DeepSeek-V3-0324 released with reasoning and function-calling upgrades","sourceUrl":"https://api-docs.deepseek.com/news/news250325","modelName":"DeepSeek-V3-0324","orgName":"DeepSeek","countryCode":"CN"},{"id":"7269f91b-b3cb-4656-b3f4-e4f26098aff5","modelSlug":"mistral-small-3-1","eventType":"released","eventDate":"2025-03-17","title":"Mistral releases Mistral Small 3.1","sourceUrl":"https://mistral.ai/news/mistral-small-3-1/","modelName":"Mistral Small 3.1","orgName":"Mistral AI","countryCode":"FR"},{"id":"d792d82a-6046-4976-9d54-b31af01bb982","modelSlug":"ernie-x1","eventType":"released","eventDate":"2025-03-16","title":"Baidu releases ERNIE X1","sourceUrl":"https://yiyan.baidu.com/","modelName":"ERNIE X1","orgName":"Baidu","countryCode":"CN"},{"id":"8f550467-fdb6-4140-86d4-0f9d46c6ffe9","modelSlug":"cohere-command-a","eventType":"released","eventDate":"2025-03-13","title":"Cohere releases Command A","sourceUrl":"https://cohere.com","modelName":"Command A","orgName":"Cohere","countryCode":"CA"},{"id":"449d174b-e1aa-4e22-bdca-77b5c7fe99ca","modelSlug":"olmo-2-32b","eventType":"released","eventDate":"2025-03-13","title":"Ai2 releases OLMo 2 32B — fully open weights, data, and code","sourceUrl":"https://allenai.org/blog/olmo2-32B","modelName":"OLMo 2 32B","orgName":"Allen Institute for AI (Ai2)","countryCode":"US"},{"id":"ae8ca85e-9a2f-4b23-8658-1dcfa46836b4","modelSlug":"granite-3-2-8b","eventType":"released","eventDate":"2025-02-26","title":"IBM releases Granite 3.2 8B","sourceUrl":"https://www.ibm.com/new/announcements/ibm-granite-3-2-expands-enterprise-ai-options","modelName":"Granite 3.2 8B","orgName":"IBM","countryCode":"US"},{"id":"ba30c41a-7468-4335-9757-e54d9f65c25f","modelSlug":"moonlight-16b-a3b-instruct","eventType":"released","eventDate":"2025-02-24","title":"Moonshot releases Moonlight-16B-A3B-Instruct","summary":"Open 16B/3B-active MoE demonstrating Moonshot's scalable Muon optimizer work.","sourceUrl":"https://huggingface.co/moonshotai/Moonlight-16B-A3B-Instruct","modelName":"Moonlight-16B-A3B-Instruct","orgName":"Moonshot AI","countryCode":"CN"},{"id":"3a22f43f-99f4-4867-bcbc-77a24efc69b2","modelSlug":"deephermes-3-llama-3-8b","eventType":"released","eventDate":"2025-02-18","title":"Nous Research releases DeepHermes 3 Llama 3 8B","sourceUrl":"https://huggingface.co/NousResearch/DeepHermes-3-Llama-3-8B-Preview","modelName":"DeepHermes 3 Llama 3 8B","orgName":"Nous Research","countryCode":"US"},{"id":"1de8954d-d3bf-41d2-846c-d82d5f0b6b5a","modelSlug":"grok-3","eventType":"released","eventDate":"2025-02-17","title":"xAI releases Grok 3","sourceUrl":"https://x.ai/news/grok-3","modelName":"Grok 3","orgName":"xAI","countryCode":"US"},{"id":"73ab275e-6604-4e39-b963-1a956dc4d798","modelSlug":"dolphin-3-0-llama3-1-8b","eventType":"released","eventDate":"2025-02-02","title":"Cognitive Computations releases Dolphin 3.0 Llama 3.1 8B","sourceUrl":"https://huggingface.co/dphn/Dolphin3.0-Llama3.1-8B","modelName":"Dolphin 3.0 Llama 3.1 8B","orgName":"Cognitive Computations","countryCode":"US"},{"id":"22c4df2e-e1cb-4930-89a4-3538409a2bb5","modelSlug":"mistral-small-3","eventType":"released","eventDate":"2025-01-30","title":"Mistral Small 3 (24B, Apache-2.0) released","sourceUrl":"https://mistral.ai/news/mistral-small-3/","modelName":"Mistral Small 3","orgName":"Mistral AI","countryCode":"FR"},{"id":"6ab987cc-0bd8-47ae-9d63-1992537a47ba","modelSlug":"qwen2-5-max","eventType":"released","eventDate":"2025-01-29","title":"Qwen releases Qwen2.5-Max","sourceUrl":"https://qwenlm.github.io/blog/qwen2.5-max/","modelName":"Qwen2.5-Max","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"1d96f395-6b54-438d-b3e5-94264d111612","modelSlug":"qwen2-5-vl-72b","eventType":"released","eventDate":"2025-01-26","title":"Qwen releases Qwen2.5-VL-72B","sourceUrl":"https://qwenlm.github.io/blog/qwen2.5-vl/","modelName":"Qwen2.5-VL-72B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"6506cf16-2373-4325-b64e-c8aaeedd2e47","modelSlug":"doubao-1-5-pro","eventType":"released","eventDate":"2025-01-22","title":"ByteDance Seed releases Doubao-1.5-pro","sourceUrl":"https://www.volcengine.com/product/doubao","modelName":"Doubao-1.5-pro","orgName":"ByteDance Seed","countryCode":"CN"},{"id":"7388114c-7f7e-4118-b7d2-654924d6c4be","modelSlug":"kimi-k1-5","eventType":"released","eventDate":"2025-01-20","title":"Moonshot releases Kimi k1.5","summary":"Multimodal reinforcement-learning reasoning model reported to match OpenAI o1 on math, coding, and multimodal reasoning.","sourceUrl":"https://arxiv.org/abs/2501.12599","modelName":"Kimi k1.5","orgName":"Moonshot AI","countryCode":"CN"},{"id":"6fb08760-de8d-4e61-bef5-ccd720cfccad","modelSlug":"deepseek-r1","eventType":"released","eventDate":"2025-01-20","title":"DeepSeek-R1 released under MIT","sourceUrl":"https://api-docs.deepseek.com/news/news250120","modelName":"DeepSeek-R1","orgName":"DeepSeek","countryCode":"CN"},{"id":"35afb215-ba89-443c-88bd-bbecb705480c","modelSlug":"minimax-01","eventType":"released","eventDate":"2025-01-15","title":"MiniMax releases MiniMax-01","sourceUrl":"https://huggingface.co/MiniMaxAI/MiniMax-Text-01","modelName":"MiniMax-01","orgName":"MiniMax","countryCode":"CN"},{"id":"4c5c0736-8fb3-4398-b7a7-4c18bc12e926","modelSlug":"deepseek-v3","eventType":"released","eventDate":"2024-12-26","title":"DeepSeek-V3 released as a 671B/37B-active MoE","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V3","modelName":"DeepSeek-V3","orgName":"DeepSeek","countryCode":"CN"},{"id":"7df81702-d487-4265-9ea6-ce82f6bff35d","modelSlug":"step-2","eventType":"released","eventDate":"2024-12-23","title":"StepFun releases Step-2","sourceUrl":"https://www.stepfun.com/","modelName":"Step-2","orgName":"StepFun","countryCode":"CN"},{"id":"23aca1f6-4a00-41f2-a2f0-28655bd75fb4","modelSlug":"granite-3-1-8b","eventType":"released","eventDate":"2024-12-18","title":"IBM releases Granite 3.1 with 128k context","sourceUrl":"https://www.ibm.com/new/announcements/ibm-granite-3-1-powerful-performance-long-context-and-more","modelName":"Granite 3.1 8B","orgName":"IBM","countryCode":"US"},{"id":"a0505fc3-1b00-4790-92c9-2af98e42da33","modelSlug":"falcon-3-10b","eventType":"released","eventDate":"2024-12-17","title":"TII releases the Falcon 3 small-model family","sourceUrl":"https://www.tii.ae/news/falcon-3-uaes-technology-innovation-institute-launches-worlds-most-powerful-small-ai-models","modelName":"Falcon 3 10B","orgName":"Technology Innovation Institute","countryCode":"AE"},{"id":"aed3d7fd-2e2b-4544-abcf-d314ab44be91","modelSlug":"command-r7b","eventType":"released","eventDate":"2024-12-13","title":"Cohere releases Command R7B","sourceUrl":"https://docs.cohere.com/changelog/command-r-7b","modelName":"Command R7B","orgName":"Cohere","countryCode":"CA"},{"id":"ef339af6-791d-4ba4-bae5-5735d21ea0c2","modelSlug":"phi-4","eventType":"released","eventDate":"2024-12-12","title":"Microsoft releases Phi-4 (14B) under MIT","sourceUrl":"https://azure.microsoft.com/products/phi","modelName":"Phi-4","orgName":"Microsoft","countryCode":"US"},{"id":"d8da9b52-e34a-4684-a916-e677cbbab54a","modelSlug":"gemini-2-0-flash","eventType":"released","eventDate":"2024-12-11","title":"Google DeepMind releases Gemini 2.0 Flash","sourceUrl":"https://blog.google/technology/google-deepmind/google-gemini-ai-update-december-2024/","modelName":"Gemini 2.0 Flash","orgName":"Google DeepMind","countryCode":"US"},{"id":"8348a58b-918c-45fc-b25e-bbf5b91dfd09","modelSlug":"exaone-3-5-32b","eventType":"released","eventDate":"2024-12-09","title":"LG AI Research releases EXAONE 3.5 32B","sourceUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-3.5-32B-Instruct","modelName":"EXAONE 3.5 32B","orgName":"LG AI Research","countryCode":"KR"},{"id":"5d412f2a-752c-4d3b-8eff-6dff3079a80f","modelSlug":"llama-3-3-70b","eventType":"released","eventDate":"2024-12-06","title":"Meta releases Llama 3.3 70B","sourceUrl":"https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct","modelName":"Llama 3.3 70B","orgName":"Meta AI","countryCode":"US"},{"id":"34764a5a-939f-4465-9306-9bd924754b2f","modelSlug":"o1","eventType":"released","eventDate":"2024-12-05","title":"OpenAI releases OpenAI o1","sourceUrl":"https://openai.com/index/introducing-openai-o1/","modelName":"OpenAI o1","orgName":"OpenAI","countryCode":"US"},{"id":"90031a19-d421-4493-9187-1663685c3fdf","modelSlug":"amazon-nova-lite","eventType":"released","eventDate":"2024-12-03","title":"Amazon releases Amazon Nova Lite","sourceUrl":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","modelName":"Amazon Nova Lite","orgName":"Amazon","countryCode":"US"},{"id":"6b707de3-3f0d-4f64-b9ce-f975b904ca59","modelSlug":"amazon-nova-pro","eventType":"released","eventDate":"2024-12-03","title":"Amazon introduces the Nova foundation models","sourceUrl":"https://aws.amazon.com/blogs/aws/introducing-amazon-nova-frontier-intelligence-and-industry-leading-price-performance/","modelName":"Amazon Nova Pro","orgName":"Amazon","countryCode":"US"},{"id":"6808f2be-4644-49df-a861-f123a4a81fca","modelSlug":"qwq-32b-preview","eventType":"released","eventDate":"2024-11-28","title":"Qwen releases QwQ-32B-Preview","sourceUrl":"https://qwenlm.github.io/blog/qwq-32b-preview/","modelName":"QwQ-32B-Preview","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"637ae3fc-5a02-4aa9-8867-1decdfbf40e9","modelSlug":"tulu-3-405b","eventType":"released","eventDate":"2024-11-21","title":"Ai2 releases Tulu 3 405B","sourceUrl":"https://allenai.org/blog/tulu-3","modelName":"Tulu 3 405B","orgName":"Allen Institute for AI (Ai2)","countryCode":"US"},{"id":"b07df274-3a07-4356-96d4-f1edc893de09","modelSlug":"deepseek-r1-lite-preview","eventType":"released","eventDate":"2024-11-20","title":"DeepSeek releases DeepSeek-R1-Lite-Preview","sourceUrl":"https://api-docs.deepseek.com/news/news1120","modelName":"DeepSeek-R1-Lite-Preview","orgName":"DeepSeek","countryCode":"CN"},{"id":"7a063324-ffba-4075-afd7-1e0f8027b041","modelSlug":"qwen2-5-coder-32b","eventType":"released","eventDate":"2024-11-12","title":"Qwen releases Qwen2.5-Coder-32B","sourceUrl":"https://qwenlm.github.io/blog/qwen2.5-coder-family/","modelName":"Qwen2.5-Coder-32B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"6e0e3e00-5236-4ecb-9da3-5c59063ebad9","modelSlug":"smollm2-1-7b","eventType":"released","eventDate":"2024-11-04","title":"Hugging Face releases SmolLM2 1.7B","sourceUrl":"https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct","modelName":"SmolLM2 1.7B","orgName":"Hugging Face","countryCode":"US"},{"id":"d8ca67c2-946d-4a40-9bee-6fc6c3c379e9","modelSlug":"hunyuan-large","eventType":"released","eventDate":"2024-11-04","title":"Tencent releases Hunyuan-Large as a 389B/52B-active MoE","sourceUrl":"https://huggingface.co/tencent/Tencent-Hunyuan-Large","modelName":"Hunyuan-Large","orgName":"Tencent Hunyuan","countryCode":"CN"},{"id":"c5a2254a-1cd5-4143-8f9c-fed47ae61515","modelSlug":"claude-3-5-haiku","eventType":"released","eventDate":"2024-10-22","title":"Anthropic releases Claude 3.5 Haiku","sourceUrl":"https://www.anthropic.com/news/3-5-models-and-computer-use","modelName":"Claude 3.5 Haiku","orgName":"Anthropic","countryCode":"US"},{"id":"64e7dd3c-a1d8-4857-a60b-7c813dfa65f2","modelSlug":"sarvam-1","eventType":"released","eventDate":"2024-10-22","title":"Sarvam AI releases Sarvam-1","sourceUrl":"https://www.sarvam.ai/blogs/sarvam-1","modelName":"Sarvam-1","orgName":"Sarvam AI","countryCode":"IN"},{"id":"54856fa3-831e-428e-a76f-fd2d427ecfda","modelSlug":"granite-3-0-8b","eventType":"released","eventDate":"2024-10-21","title":"IBM releases Granite 3.0 8B","sourceUrl":"https://www.ibm.com/new/announcements/ibm-granite-3-0-open-state-of-the-art-enterprise-models","modelName":"Granite 3.0 8B","orgName":"IBM","countryCode":"US"},{"id":"e20aef4c-18ed-427a-95cd-8e3ff2cfb013","modelSlug":"ministral-8b","eventType":"released","eventDate":"2024-10-16","title":"Mistral releases Ministral 8B","sourceUrl":"https://mistral.ai/news/ministraux/","modelName":"Ministral 8B","orgName":"Mistral AI","countryCode":"FR"},{"id":"f980c979-1267-4468-8be2-e93a6e45b5b4","modelSlug":"yi-lightning","eventType":"released","eventDate":"2024-10-16","title":"01.AI launches Yi-Lightning, reaching the Arena top-10","sourceUrl":"https://www.aifire.co/p/yi-lightning-ai-model-by-china-ranks-6-in-lmsys-arena","modelName":"Yi-Lightning","orgName":"01.AI","countryCode":"CN"},{"id":"4943cc71-cd83-4981-9be0-64b405cc4772","modelSlug":"llama-3-1-nemotron-70b","eventType":"released","eventDate":"2024-10-15","title":"NVIDIA releases Llama-3.1-Nemotron-70B","sourceUrl":"https://huggingface.co/nvidia/Llama-3.1-Nemotron-70B-Instruct-HF","modelName":"Llama-3.1-Nemotron-70B","orgName":"NVIDIA","countryCode":"US"},{"id":"bbd800f5-4d11-4677-867e-0a88bbf7aa5b","modelSlug":"llama-3-2-90b-vision","eventType":"released","eventDate":"2024-09-25","title":"Meta releases Llama 3.2 90B Vision","sourceUrl":"https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/","modelName":"Llama 3.2 90B Vision","orgName":"Meta AI","countryCode":"US"},{"id":"9a32007d-78e0-4154-aac0-730b2b7caca7","modelSlug":"molmo-72b","eventType":"released","eventDate":"2024-09-25","title":"Ai2 releases Molmo 72B","sourceUrl":"https://molmo.allenai.org/","modelName":"Molmo 72B","orgName":"Allen Institute for AI (Ai2)","countryCode":"US"},{"id":"b891c462-70e6-4f9b-a8d5-b105384228a1","modelSlug":"qwen2-5-72b","eventType":"released","eventDate":"2024-09-19","title":"Qwen releases Qwen2.5-72B","sourceUrl":"https://qwenlm.github.io/blog/qwen2.5/","modelName":"Qwen2.5-72B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"89bb8675-d2b3-43cd-886c-8158f71a0a8f","modelSlug":"pixtral-12b","eventType":"released","eventDate":"2024-09-17","title":"Mistral releases Pixtral 12B","sourceUrl":"https://mistral.ai/news/pixtral-12b/","modelName":"Pixtral 12B","orgName":"Mistral AI","countryCode":"FR"},{"id":"3d0086a0-ab01-4937-8501-1f59b55766fb","modelSlug":"o1-preview","eventType":"released","eventDate":"2024-09-12","title":"OpenAI releases OpenAI o1-preview","sourceUrl":"https://openai.com/index/introducing-openai-o1-preview/","modelName":"OpenAI o1-preview","orgName":"OpenAI","countryCode":"US"},{"id":"16a9a39e-0af8-4a7d-b682-057e4bc519ed","modelSlug":"hunyuan-turbo","eventType":"released","eventDate":"2024-09-05","title":"Tencent Hunyuan releases Hunyuan Turbo","sourceUrl":"https://hunyuan.tencent.com/","modelName":"Hunyuan Turbo","orgName":"Tencent Hunyuan","countryCode":"CN"},{"id":"80c62ee7-c0c2-4bbd-9e02-dca0d4872fb1","modelSlug":"deepseek-v2-5","eventType":"released","eventDate":"2024-09-05","title":"DeepSeek releases DeepSeek-V2.5","sourceUrl":"https://huggingface.co/deepseek-ai/DeepSeek-V2.5","modelName":"DeepSeek-V2.5","orgName":"DeepSeek","countryCode":"CN"},{"id":"10b34d05-2b2f-4f3b-8291-a37e0d5df7a9","modelSlug":"yi-coder-9b","eventType":"released","eventDate":"2024-09-05","title":"01.AI releases Yi-Coder-9B","sourceUrl":"https://huggingface.co/01-ai/Yi-Coder-9B-Chat","modelName":"Yi-Coder-9B","orgName":"01.AI","countryCode":"CN"},{"id":"af659ade-0d7a-469a-8a08-acafb58f2991","modelSlug":"olmoe-1b-7b","eventType":"released","eventDate":"2024-09-03","title":"Ai2 releases OLMoE 1B-7B","sourceUrl":"https://allenai.org/blog/olmoe","modelName":"OLMoE 1B-7B","orgName":"Allen Institute for AI (Ai2)","countryCode":"US"},{"id":"025aa2ed-4066-4e41-9268-7b79a9b32fe7","modelSlug":"jamba-1-5-large","eventType":"released","eventDate":"2024-08-22","title":"AI21 releases Jamba 1.5 hybrid Mamba-Transformer models","sourceUrl":"https://www.ai21.com/research/jamba-1-5-hybrid-transformer-mamba-models-at-scale/","modelName":"Jamba 1.5 Large","orgName":"AI21 Labs","countryCode":"IL"},{"id":"5b963d0f-b249-4fae-ad85-34e863ce895d","modelSlug":"phi-3-5-moe","eventType":"released","eventDate":"2024-08-20","title":"Microsoft releases Phi-3.5 MoE","sourceUrl":"https://huggingface.co/microsoft/Phi-3.5-MoE-instruct","modelName":"Phi-3.5 MoE","orgName":"Microsoft","countryCode":"US"},{"id":"8a4f073d-df8b-445e-b5cf-c441724bd0c8","modelSlug":"hermes-3-llama-3-1-405b","eventType":"released","eventDate":"2024-08-15","title":"Nous Research releases Hermes 3 Llama 3.1 405B","sourceUrl":"https://huggingface.co/NousResearch/Hermes-3-Llama-3.1-405B","modelName":"Hermes 3 Llama 3.1 405B","orgName":"Nous Research","countryCode":"US"},{"id":"012e3e14-22f9-419c-b0c9-d01c707ca4d1","modelSlug":"grok-2","eventType":"released","eventDate":"2024-08-13","title":"xAI releases Grok-2","sourceUrl":"https://x.ai/news/grok-2","modelName":"Grok-2","orgName":"xAI","countryCode":"US"},{"id":"52ac9ebc-02c1-42bd-8c10-11fdbdfb7327","modelSlug":"exaone-3-0-7-8b","eventType":"released","eventDate":"2024-08-07","title":"LG AI Research releases EXAONE 3.0 7.8B","sourceUrl":"https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct","modelName":"EXAONE 3.0 7.8B","orgName":"LG AI Research","countryCode":"KR"},{"id":"9c8ca1e6-3db8-4a6d-8409-4edc072087c4","modelSlug":"minicpm-v-2-6","eventType":"released","eventDate":"2024-08-02","title":"OpenBMB releases MiniCPM-V 2.6","sourceUrl":"https://huggingface.co/openbmb/MiniCPM-V-2_6","modelName":"MiniCPM-V 2.6","orgName":"OpenBMB","countryCode":"CN"},{"id":"9db8bcd4-d3c5-4edd-8ce0-59892fcf01db","modelSlug":"llama-3-1-405b","eventType":"released","eventDate":"2024-07-23","title":"Meta releases Llama 3.1 405B","sourceUrl":"https://ai.meta.com/blog/meta-llama-3-1/","modelName":"Llama 3.1 405B","orgName":"Meta AI","countryCode":"US"},{"id":"f13f6586-962e-4f01-9b8d-c19f0c4fb6c1","modelSlug":"mistral-nemo","eventType":"released","eventDate":"2024-07-18","title":"Mistral releases Mistral NeMo","sourceUrl":"https://mistral.ai/news/mistral-nemo/","modelName":"Mistral NeMo","orgName":"Mistral AI","countryCode":"FR"},{"id":"5ce819e7-19c5-4545-ae58-499a3725753e","modelSlug":"gemma-2-27b","eventType":"released","eventDate":"2024-06-27","title":"Google DeepMind releases Gemma 2 27B","sourceUrl":"https://blog.google/technology/developers/google-gemma-2/","modelName":"Gemma 2 27B","orgName":"Google DeepMind","countryCode":"US"},{"id":"00dd8e35-d657-42ff-bbf4-faabf424bf8f","modelSlug":"claude-3-5-sonnet-20240620","eventType":"released","eventDate":"2024-06-20","title":"Anthropic releases Claude 3.5 Sonnet","sourceUrl":"https://www.anthropic.com/news/claude-3-5-sonnet","modelName":"Claude 3.5 Sonnet","orgName":"Anthropic","countryCode":"US"},{"id":"af0fb0b7-1bef-4895-8f2a-db2fc28bd766","modelSlug":"deepseek-coder-v2","eventType":"released","eventDate":"2024-06-17","title":"DeepSeek releases DeepSeek-Coder-V2","sourceUrl":"https://github.com/deepseek-ai/DeepSeek-Coder-V2","modelName":"DeepSeek-Coder-V2","orgName":"DeepSeek","countryCode":"CN"},{"id":"7853cf36-f261-4d9d-89ee-0c47495eb7b4","modelSlug":"nemotron-4-340b","eventType":"released","eventDate":"2024-06-14","title":"NVIDIA releases Nemotron-4 340B","sourceUrl":"https://huggingface.co/nvidia/Nemotron-4-340B-Instruct","modelName":"Nemotron-4 340B","orgName":"NVIDIA","countryCode":"US"},{"id":"bd0e6b71-d7d0-456a-99d7-c6a41fa84494","modelSlug":"qwen2-72b","eventType":"released","eventDate":"2024-06-07","title":"Qwen releases Qwen2-72B","sourceUrl":"https://qwenlm.github.io/blog/qwen2/","modelName":"Qwen2-72B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"a629488b-ae89-409d-996d-ebbea0774a8f","modelSlug":"glm-4-9b","eventType":"released","eventDate":"2024-06-05","title":"Zhipu AI releases GLM-4-9B","sourceUrl":"https://github.com/THUDM/GLM-4","modelName":"GLM-4-9B","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"f0c96281-ecbb-4d82-9da6-303402acfa85","modelSlug":"codestral-22b","eventType":"released","eventDate":"2024-05-29","title":"Mistral releases Codestral 22B","sourceUrl":"https://mistral.ai/news/codestral/","modelName":"Codestral 22B","orgName":"Mistral AI","countryCode":"FR"},{"id":"a97deb2b-f1f3-4eff-9561-aa245bd8e306","modelSlug":"aya-23-35b","eventType":"released","eventDate":"2024-05-23","title":"Cohere releases Aya 23 35B","sourceUrl":"https://cohere.com/research/aya","modelName":"Aya 23 35B","orgName":"Cohere","countryCode":"CA"},{"id":"35e40095-4111-4693-8371-df1f263a97dc","modelSlug":"doubao-pro","eventType":"released","eventDate":"2024-05-15","title":"ByteDance Seed releases Doubao-pro","sourceUrl":"https://www.volcengine.com/product/doubao","modelName":"Doubao-pro","orgName":"ByteDance Seed","countryCode":"CN"},{"id":"df7717c8-ffff-4e85-ba6d-08502cae1cc8","modelSlug":"yi-1-5-34b","eventType":"released","eventDate":"2024-05-13","title":"01.AI releases Yi-1.5-34B","sourceUrl":"https://huggingface.co/01-ai/Yi-1.5-34B-Chat","modelName":"Yi-1.5-34B","orgName":"01.AI","countryCode":"CN"},{"id":"58613d15-826e-404a-87a5-1097e9109534","modelSlug":"falcon-2-11b","eventType":"released","eventDate":"2024-05-13","title":"TII releases Falcon 2 11B","sourceUrl":"https://www.tii.ae/news/technology-innovation-institute-launches-falcon-2-series","modelName":"Falcon 2 11B","orgName":"Technology Innovation Institute","countryCode":"AE"},{"id":"a345c796-de51-4c8a-b586-0132fbcf800d","modelSlug":"deepseek-v2","eventType":"released","eventDate":"2024-05-07","title":"DeepSeek releases DeepSeek-V2","sourceUrl":"https://github.com/deepseek-ai/DeepSeek-V2","modelName":"DeepSeek-V2","orgName":"DeepSeek","countryCode":"CN"},{"id":"2e6defc5-5a8a-45ac-90a0-b804e8679afd","modelSlug":"granite-code-34b","eventType":"released","eventDate":"2024-05-06","title":"IBM releases Granite Code 34B","sourceUrl":"https://huggingface.co/ibm-granite/granite-34b-code-instruct","modelName":"Granite Code 34B","orgName":"IBM","countryCode":"US"},{"id":"30b5faba-c4dc-48a7-a917-fe2f4f3406e6","modelSlug":"amazon-titan-text-premier","eventType":"released","eventDate":"2024-04-30","title":"Amazon releases Amazon Titan Text Premier","sourceUrl":"https://aws.amazon.com/blogs/aws/amazon-titan-text-premier-is-now-available-in-amazon-bedrock/","modelName":"Amazon Titan Text Premier","orgName":"Amazon","countryCode":"US"},{"id":"46ef1df7-7b6c-470d-bc97-4c515afa43ad","modelSlug":"snowflake-arctic","eventType":"released","eventDate":"2024-04-24","title":"Snowflake releases Snowflake Arctic","sourceUrl":"https://www.snowflake.com/en/blog/arctic-open-efficient-foundation-language-models-snowflake/","modelName":"Snowflake Arctic","orgName":"Snowflake AI Research","countryCode":"US"},{"id":"370c6321-3b5a-46f3-9eea-bf0b650e1d69","modelSlug":"phi-3-mini","eventType":"released","eventDate":"2024-04-23","title":"Microsoft releases Phi-3 Mini","sourceUrl":"https://www.microsoft.com/en-us/research/publication/phi-3-technical-report-a-highly-capable-language-model-locally-on-your-phone/","modelName":"Phi-3 Mini","orgName":"Microsoft","countryCode":"US"},{"id":"6b1b6fe9-ee78-4efb-a4dc-63a34a450f64","modelSlug":"llama-3-70b","eventType":"released","eventDate":"2024-04-18","title":"Meta releases Llama 3 70B","sourceUrl":"https://ai.meta.com/blog/meta-llama-3/","modelName":"Llama 3 70B","orgName":"Meta AI","countryCode":"US"},{"id":"d69cb930-3148-4db7-9466-67b0ab4c7803","modelSlug":"minimax-abab6-5","eventType":"released","eventDate":"2024-04-17","title":"MiniMax releases abab6.5","sourceUrl":"https://www.minimax.io/","modelName":"abab6.5","orgName":"MiniMax","countryCode":"CN"},{"id":"19f45776-ae90-4c92-9c91-66318b61386a","modelSlug":"mixtral-8x22b","eventType":"released","eventDate":"2024-04-17","title":"Mistral releases Mixtral 8x22B","sourceUrl":"https://mistral.ai/news/mixtral-8x22b/","modelName":"Mixtral 8x22B","orgName":"Mistral AI","countryCode":"FR"},{"id":"e02f8d20-917e-4ce7-a8f5-47dabb0e722f","modelSlug":"wizardlm-2-8x22b","eventType":"released","eventDate":"2024-04-15","title":"Microsoft releases WizardLM-2 8x22B","sourceUrl":"https://huggingface.co/microsoft/WizardLM-2-8x22B","modelName":"WizardLM-2 8x22B","orgName":"Microsoft","countryCode":"US"},{"id":"d103da90-3ee0-40f2-ad5c-9995d34ee490","modelSlug":"step-1v","eventType":"released","eventDate":"2024-04-12","title":"StepFun releases Step-1V","sourceUrl":"https://www.stepfun.com/","modelName":"Step-1V","orgName":"StepFun","countryCode":"CN"},{"id":"d161c252-3d6b-4d89-b94d-e2264e72f2bf","modelSlug":"codegemma-7b","eventType":"released","eventDate":"2024-04-09","title":"Google DeepMind releases CodeGemma 7B","sourceUrl":"https://developers.googleblog.com/en/codegemma-open-code-models/","modelName":"CodeGemma 7B","orgName":"Google DeepMind","countryCode":"US"},{"id":"a85b773f-9a59-4bdb-a3dd-1cfe21b8ecb4","modelSlug":"command-r-plus","eventType":"released","eventDate":"2024-04-04","title":"Cohere releases Command R+","sourceUrl":"https://cohere.com/blog/command-r-plus","modelName":"Command R+","orgName":"Cohere","countryCode":"CA"},{"id":"3bdd21b8-5a4b-4bee-9629-bb67041bfa24","modelSlug":"grok-1-5","eventType":"released","eventDate":"2024-03-28","title":"xAI releases Grok-1.5","sourceUrl":"https://x.ai/news/grok-1.5","modelName":"Grok-1.5","orgName":"xAI","countryCode":"US"},{"id":"10070153-8b39-42a9-b78f-0d8f32ae3639","modelSlug":"jamba","eventType":"released","eventDate":"2024-03-28","title":"AI21 releases Jamba","sourceUrl":"https://www.ai21.com/blog/announcing-jamba","modelName":"Jamba","orgName":"AI21 Labs","countryCode":"IL"},{"id":"3d48f76d-0beb-4eca-957c-064d650ef785","modelSlug":"dbrx-instruct","eventType":"released","eventDate":"2024-03-27","title":"Databricks releases DBRX Instruct","sourceUrl":"https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm","modelName":"DBRX Instruct","orgName":"Databricks / MosaicML","countryCode":"US"},{"id":"a2f3c701-d1df-4f7d-a7cd-f3feb802108d","modelSlug":"step-1","eventType":"released","eventDate":"2024-03-23","title":"StepFun releases Step-1","sourceUrl":"https://www.stepfun.com/","modelName":"Step-1","orgName":"StepFun","countryCode":"CN"},{"id":"67545724-81a1-4385-ac25-4bd5ea1b0bc4","modelSlug":"kimi-1m","eventType":"released","eventDate":"2024-03-18","title":"Moonshot releases Kimi 1M","sourceUrl":"https://kimi.moonshot.cn/","modelName":"Kimi 1M","orgName":"Moonshot AI","countryCode":"CN"},{"id":"1f6275ca-ea82-4917-87df-0c7c6e659348","modelSlug":"command-r","eventType":"released","eventDate":"2024-03-11","title":"Cohere releases Command R","sourceUrl":"https://cohere.com/blog/command-r","modelName":"Command R","orgName":"Cohere","countryCode":"CA"},{"id":"d3cd397e-79e0-4104-b6c9-4c163f26dbf3","modelSlug":"claude-3-opus","eventType":"released","eventDate":"2024-03-04","title":"Anthropic releases Claude 3 Opus","sourceUrl":"https://www.anthropic.com/news/claude-3-family","modelName":"Claude 3 Opus","orgName":"Anthropic","countryCode":"US"},{"id":"1b54f51d-caba-4085-94ce-4fd6825e3bb3","modelSlug":"starcoder2-15b","eventType":"released","eventDate":"2024-02-28","title":"BigCode releases StarCoder2 15B","sourceUrl":"https://huggingface.co/blog/starcoder2","modelName":"StarCoder2 15B","orgName":"BigCode","countryCode":"US"},{"id":"e3a2b2e7-6c14-484e-a392-50ce152d9262","modelSlug":"mistral-large","eventType":"released","eventDate":"2024-02-26","title":"Mistral releases Mistral Large","sourceUrl":"https://mistral.ai/news/mistral-large/","modelName":"Mistral Large","orgName":"Mistral AI","countryCode":"FR"},{"id":"0bc0507a-df7d-41eb-8bbd-82762f8537b6","modelSlug":"gemma-7b","eventType":"released","eventDate":"2024-02-21","title":"Google DeepMind releases Gemma 7B","sourceUrl":"https://blog.google/technology/developers/gemma-open-models/","modelName":"Gemma 7B","orgName":"Google DeepMind","countryCode":"US"},{"id":"76d341bf-9613-4a33-a28d-fd8d852d3446","modelSlug":"gemini-1-5-pro","eventType":"released","eventDate":"2024-02-15","title":"Google DeepMind releases Gemini 1.5 Pro","sourceUrl":"https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/","modelName":"Gemini 1.5 Pro","orgName":"Google DeepMind","countryCode":"US"},{"id":"be689e40-13e0-4494-8d6f-c08220824afa","modelSlug":"qwen1-5-110b","eventType":"released","eventDate":"2024-02-05","title":"Qwen releases Qwen1.5-110B","sourceUrl":"https://qwenlm.github.io/blog/qwen1.5/","modelName":"Qwen1.5-110B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"3b344b06-7964-4790-9c11-be900621039d","modelSlug":"olmo-7b","eventType":"released","eventDate":"2024-02-01","title":"Ai2 releases OLMo 7B","sourceUrl":"https://allenai.org/blog/olmo-open-language-model","modelName":"OLMo 7B","orgName":"Allen Institute for AI (Ai2)","countryCode":"US"},{"id":"9d115157-ab29-4097-91ea-f097fe9702ae","modelSlug":"stablelm-2-1-6b","eventType":"released","eventDate":"2024-01-19","title":"Stability AI releases Stable LM 2 1.6B","sourceUrl":"https://stability.ai/news-updates/introducing-stable-lm-2","modelName":"Stable LM 2 1.6B","orgName":"Stability AI","countryCode":"GB"},{"id":"3390d82e-4fe5-4da5-91df-020b3c838aa3","modelSlug":"glm-4","eventType":"released","eventDate":"2024-01-16","title":"Zhipu AI releases GLM-4","sourceUrl":"https://open.bigmodel.cn/dev/howuse/model","modelName":"GLM-4","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"9812ae36-c839-45fb-b13d-946294682db1","modelSlug":"nous-hermes-2-mixtral","eventType":"released","eventDate":"2024-01-11","title":"Nous Research releases Nous Hermes 2 Mixtral","sourceUrl":"https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO","modelName":"Nous Hermes 2 Mixtral","orgName":"Nous Research","countryCode":"US"},{"id":"0b236d3e-ae77-4d1b-b566-39a07ef0d663","modelSlug":"deepseek-moe-16b","eventType":"released","eventDate":"2024-01-11","title":"DeepSeek releases DeepSeekMoE 16B","sourceUrl":"https://github.com/deepseek-ai/DeepSeek-MoE","modelName":"DeepSeekMoE 16B","orgName":"DeepSeek","countryCode":"CN"},{"id":"d116cb38-c8f8-4fb6-830e-0ac68e379a8a","modelSlug":"openchat-3-5","eventType":"released","eventDate":"2024-01-06","title":"OpenChat releases OpenChat 3.5","sourceUrl":"https://huggingface.co/openchat/openchat-3.5-0106","modelName":"OpenChat 3.5","orgName":"OpenChat","countryCode":"US"},{"id":"720eb786-effa-4682-b22c-d3b61d0cda57","modelSlug":"tinyllama-1-1b-chat","eventType":"released","eventDate":"2024-01-01","title":"TinyLlama releases TinyLlama 1.1B Chat","sourceUrl":"https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0","modelName":"TinyLlama 1.1B Chat","orgName":"TinyLlama","countryCode":"US"},{"id":"3eea63f9-447f-46c2-bae8-73c093c8a4c4","modelSlug":"phi-2","eventType":"released","eventDate":"2023-12-12","title":"Microsoft releases Phi-2","sourceUrl":"https://www.microsoft.com/en-us/research/blog/phi-2-the-surprising-power-of-small-language-models/","modelName":"Phi-2","orgName":"Microsoft","countryCode":"US"},{"id":"3b8718c3-58c5-4b95-88a2-57f4a9ab3fcb","modelSlug":"openhathi-7b","eventType":"released","eventDate":"2023-12-12","title":"Sarvam AI releases OpenHathi-7B","sourceUrl":"https://www.sarvam.ai/blogs/announcing-openhathi","modelName":"OpenHathi-7B","orgName":"Sarvam AI","countryCode":"IN"},{"id":"f1d8f043-1c59-48b3-93d7-94427188624d","modelSlug":"mixtral-8x7b","eventType":"released","eventDate":"2023-12-11","title":"Mistral releases Mixtral 8x7B (open MoE)","sourceUrl":"https://mistral.ai","modelName":"Mixtral 8x7B","orgName":"Mistral AI","countryCode":"FR"},{"id":"7f6b5047-2a8b-4a3a-8a5e-1c5c4f553528","modelSlug":"gemini-1-0-ultra","eventType":"released","eventDate":"2023-12-06","title":"Google announces Gemini 1.0","sourceUrl":"https://deepmind.google","modelName":"Gemini 1.0 Ultra","orgName":"Google DeepMind","countryCode":"US"},{"id":"f9f519a1-c65e-4f3b-b2d3-a9f034cb6566","modelSlug":"qwen-72b","eventType":"released","eventDate":"2023-11-30","title":"Alibaba open-sources Qwen-72B","sourceUrl":"https://qwen.ai","modelName":"Qwen-72B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"652ef7bf-5c08-4de1-a625-c838f38d1321","modelSlug":"deepseek-llm-67b","eventType":"released","eventDate":"2023-11-29","title":"DeepSeek releases DeepSeek LLM 67B","sourceUrl":"https://github.com/deepseek-ai/DeepSeek-LLM","modelName":"DeepSeek LLM 67B","orgName":"DeepSeek","countryCode":"CN"},{"id":"d28d3950-cd27-4ca8-b64d-72c2465761aa","modelSlug":"claude-2-1","eventType":"released","eventDate":"2023-11-21","title":"Anthropic releases Claude 2.1","sourceUrl":"https://www.anthropic.com/news/claude-2-1","modelName":"Claude 2.1","orgName":"Anthropic","countryCode":"US"},{"id":"19214eca-bf6c-4c0b-ba6e-a6ae29d5aecd","modelSlug":"gpt-4-turbo","eventType":"released","eventDate":"2023-11-06","title":"OpenAI releases GPT-4 Turbo","sourceUrl":"https://openai.com/index/new-models-and-developer-products-announced-at-devday/","modelName":"GPT-4 Turbo","orgName":"OpenAI","countryCode":"US"},{"id":"8919d98f-e06e-4608-8366-c187b6301500","modelSlug":"yi-34b","eventType":"released","eventDate":"2023-11-06","title":"01.AI releases Yi-34B","sourceUrl":"https://www.01.ai","modelName":"Yi-34B","orgName":"01.AI","countryCode":"CN"},{"id":"929d6781-989c-4775-a712-3121609f63e0","modelSlug":"grok-1","eventType":"released","eventDate":"2023-11-04","title":"xAI releases Grok-1","sourceUrl":"https://x.ai/news/grok","modelName":"Grok-1","orgName":"xAI","countryCode":"US"},{"id":"69f3c12f-a854-42b5-90f1-c148ac2d1648","modelSlug":"deepseek-coder-33b","eventType":"released","eventDate":"2023-11-02","title":"DeepSeek releases DeepSeek Coder 33B","sourceUrl":"https://github.com/deepseek-ai/DeepSeek-Coder","modelName":"DeepSeek Coder 33B","orgName":"DeepSeek","countryCode":"CN"},{"id":"1b3b78b1-2b56-4647-9f99-fab6d61a2b4c","modelSlug":"ernie-4-0","eventType":"released","eventDate":"2023-10-17","title":"Baidu releases ERNIE 4.0","sourceUrl":"https://yiyan.baidu.com/","modelName":"ERNIE 4.0","orgName":"Baidu","countryCode":"CN"},{"id":"8d59e347-6e08-498f-a6b3-974cafe04792","modelSlug":"kimi-chat","eventType":"released","eventDate":"2023-10-09","title":"Moonshot releases Kimi Chat","sourceUrl":"https://kimi.moonshot.cn/","modelName":"Kimi Chat","orgName":"Moonshot AI","countryCode":"CN"},{"id":"78b31792-33ab-4e93-97d2-9c1a2b4437b0","modelSlug":"llava-1-5-13b","eventType":"released","eventDate":"2023-09-30","title":"LLaVA releases LLaVA 1.5 13B","sourceUrl":"https://llava-vl.github.io/","modelName":"LLaVA 1.5 13B","orgName":"LLaVA","countryCode":"US"},{"id":"b4507965-7bfa-48e8-b00a-ede8d78c7e00","modelSlug":"amazon-titan-text-express","eventType":"released","eventDate":"2023-09-28","title":"Amazon releases Amazon Titan Text Express","sourceUrl":"https://aws.amazon.com/blogs/aws/amazon-bedrock-is-now-generally-available-build-and-scale-generative-ai-applications-with-foundation-models/","modelName":"Amazon Titan Text Express","orgName":"Amazon","countryCode":"US"},{"id":"2e97b217-8b62-4875-ad8b-676d955521cb","modelSlug":"mistral-7b","eventType":"released","eventDate":"2023-09-27","title":"Mistral AI releases Mistral 7B","sourceUrl":"https://mistral.ai","modelName":"Mistral 7B","orgName":"Mistral AI","countryCode":"FR"},{"id":"00b84096-60c2-448c-980e-fdeec21fb8c2","modelSlug":"qwen-14b","eventType":"released","eventDate":"2023-09-25","title":"Qwen releases Qwen-14B","sourceUrl":"https://qwenlm.github.io/blog/qwen/","modelName":"Qwen-14B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"c068adee-f865-421d-9ae3-70e384c567bd","modelSlug":"hunyuan-standard","eventType":"released","eventDate":"2023-09-07","title":"Tencent Hunyuan releases Hunyuan","sourceUrl":"https://hunyuan.tencent.com/","modelName":"Hunyuan","orgName":"Tencent Hunyuan","countryCode":"CN"},{"id":"fcdf1bef-87b7-4a2f-b4e9-9afa8e478973","modelSlug":"granite-13b","eventType":"released","eventDate":"2023-09-07","title":"IBM releases Granite 13B","sourceUrl":"https://www.ibm.com/granite","modelName":"Granite 13B","orgName":"IBM","countryCode":"US"},{"id":"cbc9ea8b-e749-4088-a2ec-58242d63c7d4","modelSlug":"falcon-180b","eventType":"released","eventDate":"2023-09-06","title":"TII releases Falcon 180B","sourceUrl":"https://falconllm.tii.ae","modelName":"Falcon 180B","orgName":"Technology Innovation Institute","countryCode":"AE"},{"id":"c79701db-b752-4dda-9c6a-a994cde8e9c7","modelSlug":"code-llama-34b","eventType":"released","eventDate":"2023-08-24","title":"Meta releases Code Llama 34B","sourceUrl":"https://ai.meta.com/blog/code-llama-large-language-model-coding/","modelName":"Code Llama 34B","orgName":"Meta AI","countryCode":"US"},{"id":"6ca3425a-1417-4fb3-aa01-ddb970023578","modelSlug":"qwen-7b","eventType":"released","eventDate":"2023-08-03","title":"Qwen releases Qwen-7B","sourceUrl":"https://qwenlm.github.io/blog/qwen/","modelName":"Qwen-7B","orgName":"Alibaba (Qwen)","countryCode":"CN"},{"id":"579a2f21-7d87-4778-bb1a-75f384551b9d","modelSlug":"nous-hermes-llama2-13b","eventType":"released","eventDate":"2023-07-24","title":"Nous Research releases Nous-Hermes-Llama2-13B","sourceUrl":"https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b","modelName":"Nous-Hermes-Llama2-13B","orgName":"Nous Research","countryCode":"US"},{"id":"a065df24-d897-4ebc-aae7-6f0c762ea64c","modelSlug":"exaone-2","eventType":"released","eventDate":"2023-07-19","title":"LG AI Research releases EXAONE 2.0","sourceUrl":"https://www.lgresearch.ai/","modelName":"EXAONE 2.0","orgName":"LG AI Research","countryCode":"KR"},{"id":"5512fa8f-5a05-4a25-aba4-d2fb7e61c3dd","modelSlug":"llama-2-70b","eventType":"released","eventDate":"2023-07-18","title":"Meta releases Llama 2 under a community license","sourceUrl":"https://ai.meta.com","modelName":"Llama 2 70B","orgName":"Meta AI","countryCode":"US"},{"id":"5c41553c-c2a2-4b90-b9ad-739bd2735515","modelSlug":"claude-2","eventType":"released","eventDate":"2023-07-11","title":"Anthropic releases Claude 2 (100k context)","sourceUrl":"https://anthropic.com","modelName":"Claude 2","orgName":"Anthropic","countryCode":"US"},{"id":"79c13186-61bc-4f8e-b953-635985abf26f","modelSlug":"chatglm2-6b","eventType":"released","eventDate":"2023-06-25","title":"Zhipu AI releases ChatGLM2-6B","sourceUrl":"https://github.com/THUDM/ChatGLM2-6B","modelName":"ChatGLM2-6B","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"b875f4a6-10ed-457f-8123-b09c9525e55c","modelSlug":"phi-1","eventType":"released","eventDate":"2023-06-21","title":"Microsoft releases Phi-1","sourceUrl":"https://www.microsoft.com/en-us/research/blog/textbooks-are-all-you-need/","modelName":"Phi-1","orgName":"Microsoft","countryCode":"US"},{"id":"0e4cad47-8b36-4eaf-a69f-dcc5121a9ced","modelSlug":"falcon-40b","eventType":"released","eventDate":"2023-05-25","title":"TII releases Falcon 40B","sourceUrl":"https://falconllm.tii.ae/","modelName":"Falcon 40B","orgName":"Technology Innovation Institute","countryCode":"AE"},{"id":"e2ace67f-8343-4649-a8bf-ba6d4b4a5778","modelSlug":"palm-2","eventType":"released","eventDate":"2023-05-10","title":"Google DeepMind releases PaLM 2","sourceUrl":"https://blog.google/technology/ai/google-palm-2-ai-large-language-model/","modelName":"PaLM 2","orgName":"Google DeepMind","countryCode":"US"},{"id":"e29a3a0e-2660-47af-920f-4be35cfcd70b","modelSlug":"mpt-7b","eventType":"released","eventDate":"2023-05-05","title":"Databricks releases MPT-7B","sourceUrl":"https://www.databricks.com/blog/mpt-7b","modelName":"MPT-7B","orgName":"Databricks / MosaicML","countryCode":"US"},{"id":"590ab142-c42c-49d8-9224-8b769184bcec","modelSlug":"vicuna-13b","eventType":"released","eventDate":"2023-03-30","title":"LMSYS releases Vicuna 13B","sourceUrl":"https://lmsys.org/blog/2023-03-30-vicuna/","modelName":"Vicuna 13B","orgName":"LMSYS / SkyLab","countryCode":"US"},{"id":"6b88f156-a9f3-4a88-8324-9a8fbfd8ea25","modelSlug":"ernie-bot","eventType":"released","eventDate":"2023-03-16","title":"Baidu releases ERNIE Bot","sourceUrl":"https://yiyan.baidu.com/","modelName":"ERNIE Bot","orgName":"Baidu","countryCode":"CN"},{"id":"a79a95ce-9358-40b5-8219-b3229efbfb8b","modelSlug":"claude-1","eventType":"released","eventDate":"2023-03-14","title":"Anthropic releases Claude 1","sourceUrl":"https://www.anthropic.com/news/introducing-claude","modelName":"Claude 1","orgName":"Anthropic","countryCode":"US"},{"id":"3d385667-db71-4cb4-a282-bd351874c14b","modelSlug":"chatglm-6b","eventType":"released","eventDate":"2023-03-14","title":"Zhipu AI releases ChatGLM-6B","sourceUrl":"https://github.com/THUDM/ChatGLM-6B","modelName":"ChatGLM-6B","orgName":"Z.ai (Zhipu AI)","countryCode":"CN"},{"id":"64642acb-d7dd-40f4-a962-7a18cad9f62b","modelSlug":"gpt-4","eventType":"released","eventDate":"2023-03-14","title":"OpenAI releases GPT-4","sourceUrl":"https://openai.com/research/gpt-4","modelName":"GPT-4","orgName":"OpenAI","countryCode":"US"},{"id":"bcef90cb-a7bd-42cb-8196-5a5e1f17f7d8","modelSlug":"jurassic-2-ultra","eventType":"released","eventDate":"2023-03-09","title":"AI21 releases Jurassic-2 Ultra","sourceUrl":"https://www.ai21.com/","modelName":"Jurassic-2 Ultra","orgName":"AI21 Labs","countryCode":"IL"},{"id":"263dc42d-8253-44d6-bdf7-f47a0166b85e","modelSlug":"gpt-3-5-turbo","eventType":"released","eventDate":"2023-03-01","title":"OpenAI releases GPT-3.5 Turbo","sourceUrl":"https://openai.com/index/introducing-chatgpt-and-whisper-apis/","modelName":"GPT-3.5 Turbo","orgName":"OpenAI","countryCode":"US"},{"id":"f2252893-f94b-44db-9b7c-cb54a07b6e57","modelSlug":"llama-1","eventType":"released","eventDate":"2023-02-24","title":"Meta releases LLaMA to researchers","sourceUrl":"https://ai.meta.com","modelName":"LLaMA","orgName":"Meta AI","countryCode":"US"},{"id":"03e561f6-c0dd-45f4-b3ae-d2fbcbf0d4ec","modelSlug":"gpt-3","eventType":"updated","eventDate":"2022-11-30","title":"ChatGPT (GPT-3.5) launches and reaches 100M users","summary":"The consumer launch that ignited the modern LLM race.","sourceUrl":"https://openai.com/blog/chatgpt","modelName":"GPT-3","orgName":"OpenAI","countryCode":"US"},{"id":"258a21ce-a0c0-4b28-806f-40e000406d08","modelSlug":"galactica","eventType":"withdrawn","eventDate":"2022-11-17","title":"Meta pulls the Galactica demo after three days","summary":"Withdrawn following criticism of confident but inaccurate scientific output.","sourceUrl":"https://galactica.org","modelName":"Galactica","orgName":"Meta AI","countryCode":"US"},{"id":"d1528255-a851-49c5-bce7-0816336e0c7f","modelSlug":"galactica","eventType":"released","eventDate":"2022-11-15","title":"Meta launches the Galactica science demo","sourceUrl":"https://galactica.org","modelName":"Galactica","orgName":"Meta AI","countryCode":"US"},{"id":"1a9afc1f-a18f-4913-ba5a-16294a45f140","modelSlug":"bloom","eventType":"released","eventDate":"2022-07-12","title":"BigScience releases BLOOM (176B, multilingual)","sourceUrl":"https://bigscience.huggingface.co","modelName":"BLOOM","orgName":"BigScience","countryCode":"FR"},{"id":"4947cf0e-cf98-4e95-a127-bdbbe4610828","modelSlug":"palm","eventType":"released","eventDate":"2022-04-04","title":"Google announces PaLM (540B)","sourceUrl":"https://ai.google","modelName":"PaLM","orgName":"Google DeepMind","countryCode":"US"},{"id":"b555fc87-c473-4ed7-a9b5-6430ae1ce5c8","modelSlug":"exaone-1","eventType":"released","eventDate":"2021-12-14","title":"LG AI Research releases EXAONE 1.0","sourceUrl":"https://www.lgresearch.ai/","modelName":"EXAONE 1.0","orgName":"LG AI Research","countryCode":"KR"},{"id":"b3f8b3bc-e294-4b9a-ae26-ad79d2bc119d","modelSlug":"ernie-3-titan","eventType":"released","eventDate":"2021-12-08","title":"Baidu releases ERNIE 3.0 Titan","sourceUrl":"https://arxiv.org/abs/2112.12731","modelName":"ERNIE 3.0 Titan","orgName":"Baidu","countryCode":"CN"},{"id":"4507ef50-bcb6-4af8-8616-c45af40db731","modelSlug":"jurassic-1-jumbo","eventType":"released","eventDate":"2021-08-11","title":"AI21 releases Jurassic-1 Jumbo","sourceUrl":"https://www.ai21.com/","modelName":"Jurassic-1 Jumbo","orgName":"AI21 Labs","countryCode":"IL"},{"id":"da8d5f91-a84f-4c3e-9e86-7684042a80e3","modelSlug":"gpt-3","eventType":"released","eventDate":"2020-06-11","title":"OpenAI introduces GPT-3 (175B)","sourceUrl":"https://openai.com","modelName":"GPT-3","orgName":"OpenAI","countryCode":"US"},{"id":"7dd0b4d1-237f-46e1-896e-fa5457277255","modelSlug":"gpt-2","eventType":"released","eventDate":"2019-11-05","title":"OpenAI releases the full GPT-2 (1.5B) weights","summary":"After a staged rollout that began in Feb 2019 over misuse concerns.","sourceUrl":"https://openai.com","modelName":"GPT-2","orgName":"OpenAI","countryCode":"US"},{"id":"6b010e58-e842-4de4-a2e1-5feded35cce8","modelSlug":"bert-large","eventType":"released","eventDate":"2018-10-11","title":"Google releases BERT","sourceUrl":"https://github.com/google-research/bert","modelName":"BERT","orgName":"Google DeepMind","countryCode":"US"}]}