Mistral Launches Trillion-Param Model to Challenge China

Mistral is launching a 1-trillion-parameter open AI model claiming to lead outside China, trained on 4,000 Nvidia GPUs in Europe. The model, called Mistral Large 4 - or ML4, known as "Le Chonk" - is particularly effective at tasks related to cyber, coding, manufacturing, finance and multimodal. The model's weights will be released publicly on October 27.

Mistral unveiled the new open-weight AI model on Tuesday as Western companies race to close the capability gap with Chinese developers in the market. The model, called Mistral Large 4 or ML4, nicknamed "Le Chonk", has 1 trillion parameters and will rank among the top open-weight models globally on aggregate benchmark performance, with Mistral claiming it is the strongest open-weight model developed outside China by a substantial margin. ML4 has 49 billion active parameters and was trained from scratch over roughly two months on 4,000 Nvidia Grace Blackwell GPUs in Mistral's own European data centers. Mistral trained the model across more than 160 languages, including every official language of the European Union. On benchmark tests, ML4 scored 61.7% on DeepSWE v1.1 and 59.9% on AutomationBench; on one Artificial Analysis Cyber Index test covering vulnerability reproduction and patching, it reached 82%, while its Cybench score was 93%. The model is being launched in preview to developers and cybersecurity professionals alongside state authorities before its weights are released publicly on October 27. Mistral CEO Arthur Mensch said at a launch event that the model outperforms many rival open-weight models, stating "the narrative that Europe cannot compete is not true."