MiniMax AI has released Minimax Music 3, an open-weight music generation model that utilizes an 8B LLM and a 2.7B Diffusion Transformer. This model is capable of producing full songs from text prompts and lyrics, and is designed to run on consumer GPUs via libraries like diffusers and Comfy. Separately, MiniMax AI also announced that its H3 model, MiniMax-H3, has achieved the top rank on the Video Edit Arena, scoring 1390 points and surpassing competitors like Dreamina Seedance 2.0 and Gemini Omni Flash. AI
IMPACT Sets new SOTA for open-weight music generation and video editing benchmarks, potentially accelerating adoption of these capabilities.
RANK_REASON The cluster reports on the release of a new open-weight model (Minimax Music 3) and a top-ranking benchmark achievement (MiniMax-H3) by a frontier AI lab (MiniMax AI). [lever_c_demoted from frontier_release: ic=2 ai=1.0]
- Gemini Omni Flash
- HappyHorse 1.0
- MiniMax AI
- Video Edit Arena
- Comfy
- diffusers
- Diffusion Transformer
- Hugging Face
- Minimax Music 3
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
