Mistral releases Large 4, a 1T-parameter open-weight AI model

Mistral has unveiled a preview of Mistral Large 4, a 1-trillion-parameter open-weight model it calls the most powerful outside China. Weights arrive on October 27; training took two months on about 4,000 Nvidia Grace Blackwell GPUs in Europe. Each task uses only 49 billion of its trillion parameters.

• Model weights will be released on October 27; API access is available now
• Training ran two months on ~4,000 Grace Blackwell GPUs at ~10 MW
• Large 4 scored 62% on DeepSWE, ahead of GLM-5.3 (61%) and DeepSeek-V4-Pro (57%)
• Handles text and images, trained on more than 160 languages