Models
Models from the labs in Epoch AI's catalog, the repositories they published on Hugging Face, and posts from their own news pages.
Genie 3Google DeepMind · Games, Video, 3D modelingUnreleasedAug 5, 2025gpt-oss-120bOpenAI · Language · 117B parameters · Training compute 4.9e24 FLOPOpen weights (unrestricted)Aug 5, 2025gpt-oss-20bOpenAI · Language · 20.9B parameters · Training compute 5.5e23 FLOPOpen weights (unrestricted)Aug 5, 2025Qwen ImageAlibaba · Image generation, Vision · 27B parametersOpen weights (unrestricted)Aug 4, 2025Gemini 2.5 Deep ThinkGoogle, Google DeepMind · Language, Multimodal, Vision, Video, Audio, MathematicsHosted access (no API)Aug 1, 2025rStar-Math (Qwen2.5-Math-7B base)Microsoft Research Asia, Peking University, Tsinghua University · Language · 7B parametersUnreleasedAug 1, 2025rStar-Math (Qwen2-Math-7B base)Microsoft Research Asia, Peking University, Tsinghua University · Language · 7B parametersUnreleasedAug 1, 2025Command A VisionCohere · Vision, Multimodal, Language · 112B parametersOpen weights (non-commercial)Jul 31, 2025Veo 3 FastGoogle DeepMind · Video, VisionAPI accessJul 31, 2025AlphaEarth Foundations (AEF)Google DeepMind, Google · Earth science · 480M parameters · Training compute 8.5e21 FLOPHosted access (no API)Jul 30, 2025Llama Nemotron Super v1.5Nvidia · Language · 49B parametersOpen weights (restricted use)Jul 29, 2025Qwen3-30B-A3B (Jul 2025)Alibaba · Language · 30B parameters · Training compute 6.5e23 FLOPOpen weights (unrestricted)Jul 29, 2025GLM-4.5Z.ai (Zhipu AI), Tsinghua University · Language · 355B parameters · Training compute 4.4e24 FLOPOpen weights (unrestricted)Jul 28, 2025GLM-4.5-AirZ.ai (Zhipu AI), Tsinghua University · Language · 106B parameters · Training compute 1.7e24 FLOPOpen weights (unrestricted)Jul 28, 2025Wan 2.2 14B I2VAlibaba · Video, Vision · 14B parametersOpen weights (unrestricted)Jul 28, 2025Wan 2.2 14B T2VAlibaba · Video · 14B parameters · Training compute 4.2e23 FLOPOpen weights (unrestricted)Jul 28, 2025Qwen3-235B-A22B (Jul 2025)Alibaba · Language · 235B parameters · Training compute 4.8e24 FLOPOpen weights (unrestricted)Jul 25, 2025Qwen3-235B-A22B-Thinking (Jul 2025)Alibaba · Language · 235B parameters · Training compute 4.8e24 FLOPOpen weights (unrestricted)Jul 25, 2025Magistral Small 1.1Mistral AI · Language · 24B parametersOpen weights (unrestricted)Jul 24, 2025AeneasGoogle DeepMind, University of Nottingham, University of Warwick, Athens University of Economics and Business, Google, University of Oxford · Vision, Multimodal, Language · Training compute 2.3e21 FLOPHosted access (no API)Jul 23, 2025Qwen3-Coder-480B-A35BAlibaba · Language · 480B parameters · Training compute 1.6e24 FLOPOpen weights (unrestricted)Jul 22, 2025ChatGPT agentOpenAI · LanguageHosted access (no API)Jul 17, 2025OpenReasoning-Nemotron-32BNvidia · Language · 32B parametersOpen weights (unrestricted)Jul 16, 2025EXAONE 4.0 (1.2B)LG AI Research · Language · 1.2B parameters · Training compute 8.7e22 FLOPOpen weights (non-commercial)Jul 15, 2025EXAONE 4.0 (32B)LG AI Research · Language · 32B parameters · Training compute 2.7e24 FLOPOpen weights (non-commercial)Jul 15, 2025
Data: Epoch AI, CC BY 4.0. Changes: selected rows and columns of Epoch AI's database.