Mistral Large 4：欧洲 trillion-parameter 模型应对美国模型拒绝的安全工作

mistral has released a preview of mistral large 4, a 1-trillion-parameter model trained in european data centers, with weights expected at the end of october. it scores 38 points in the artificial analysis intelligence index, a significant leap from its predecessors, but still trails leading closed models. mistral large 4 excels in it security, reproducing and patching software vulnerabilities that competitors like claude opus 5.5 and gpt-6 refuse to handle due to safety filters. this capability is a key selling point for mistral, especially for tasks like defending software and proving vulnerabilities are real.