MiniMax M3.1 Flash 开源权重上架 Hugging Face：428B 总参、23B 激活

Citizen Oscar @oscarsuiza原推发布于 2026.10.24

在 X 上打开原推

Last Monday MiniMax launched M3.1 Flash and said open weights would come later. They just landed. The model is now on Hugging Face: about 428B parameters in total, 23B active per token, a 1M-token context window, and text, image and video input. vLLM shipped day-0 support for NVIDIA Blackwell, NVIDIA Rubin and AMD MI350 GPUs. Our take: a strong agent and coding model you can host yourself is good news for teams that can't send their data to an outside API.