Mistral launches Large 4 preview, a 1-trillion-parameter open model trained in Europe

Mistral launched a public preview of Mistral Large 4 (ML4, nicknamed “le Chonk”), a natively multimodal model with 1 trillion parameters and 49 billion active, available now via its Studio API with open weights due by the end of October. Mistral calls it the strongest open-weight model developed in the US or Europe and says it was trained from scratch on 3,800 Nvidia Grace Blackwell GPUs in its own European datacenters. Guillaume Lample says that cluster, in Bruyères-le-Châtel, was built with Mistral’s Series B money, with Series C and D clusters coming online soon. Arthur Mensch’s take: “We’re building AGI from this train station.”