News
MiniMax releases the M3 model: 1M context, native multi-modality, and leading programming capabilities. MiniMax officially releases the M3 model, which adopts the new MSA sparse attention architecture, supports up to 1 million token context windows, and can natively process images, videos, and desktop operations.
1 min read
Source: Telegram AI频道
MiniMax releases M3 model: 1M context, native multi-modality, and leading programming capabilities. MiniMax officially releases the M3 model, which adopts the new MSA sparse attention architecture, supports up to 1 million token context windows, and can natively process images, videos, and desktop operations. In the programming evaluation SWE-Bench Pro, M3 scored 59%, surpassing GPT-5.5 and Gemini 3.1 Pro; it also reached the leading level in the multi-modal OmniDocBench and Agent evaluation Claw-Eval. M3 is the first open source model in China that features ultra-long context, cutting-edge programming, and native multi-modal capabilities. …