Mistral Is SO Back: Inside Le Chonk, the West’s Biggest Open Model (Pierre Stock, VP Science, Mistral)Mistral 回来了：揭秘西方最大开放模型 Le Chonk（Mistral 科学副总裁 Pierre Stock）

Mistral 回来了：揭秘西方最大开放模型 Le Chonk（Mistral 科学副总裁 Pierre Stock）Mistral Is SO Back: Inside Le Chonk, the West’s Biggest Open Model (Pierre Stock, VP Science, Mistral)

Key points

• Mistral Large 4 (“Le Chonk”) is a 1-trillion-parameter open-weight model trained in two months on 3,800 GPUs in Mistral’s own European datacenters.
  Mistral Large 4（“Le Chonk”）是 1 万亿参数的开放权重模型，在 Mistral 自有欧洲数据中心用 3800 块 GPU 训练两个月。

• Covers what Mistral borrowed from DeepSeek, RL at scale on coding, cyber, finance and chip design, and a model that pinged its own trainer from inside the sandbox.
  讨论了 Mistral 借鉴 DeepSeek 的部分、在编程/网络安全/金融/芯片设计上的大规模 RL，以及一个在沙箱内联系自己训练者的模型。

• Stock expects 10-trillion-parameter models; Mistral Large 5 is already in pretraining.
  Stock 预计 10 万亿参数模型即将到来；Mistral Large 5 已在预训练中。