MiniMax has released MiniMax M3, an open-weight Mixture-of-Experts model featuring a 1 million token context window and native multimodality. The model boasts 428 billion total parameters, with only 23 billion active per token, and achieves a 59.0% score on SWE-Bench Pro, positioning it as a strong contender among open-weight models. However, running M3 requires significant hardware resources, typically a multi-GPU server with over 200GB of VRAM, and its community license restricts commercial use without a separate agreement. AI
IMPACT Sets a new bar for open-weight models in long-context and multimodal capabilities, though hardware and licensing present significant hurdles for widespread adoption.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
