Built for Coders,Trusted by Defenders.

Mistral has released the open weights for Mistral Large 4 (Le Chonk), a 1-trillion-parameter multimodal model. The model demonstrates leading performance on coding and defensive cybersecurity benchmarks, with significant improvements driven by reinforcement learning. It offers three distinct reasoning modes, allowing developers to balance performance and inference costs for various tasks. As a European-trained model, it provides a trusted foundation for Sovereign AI, enabling organizations to own, customize, and control their AI deployments. The model also features strong safety guardrails, effectively distinguishing between defensive and malicious cybersecurity requests.