Mistral 推出 Large 4 公开预览，称可与中国领先开放模型竞争

Mistral releases 1 trillion parameter model Mistral Large 4 (codename le Chonk), claiming its overall benchmark performance will rank among the top open-weight models globally, and that it is the strongest open-weight model "outside China", but it still lags frontier levels in areas like coding. The model was trained for two months on 4,000 Nvidia Grace Blackwell GPUs in Mistral's own European data centers, and is available in preview to developers, cybersecurity leads, and national agencies, with broader release later this month. Mistral says its cyber defense capabilities can help enterprises defend against threat actors exploiting jailbroken closed-source models.