Mistral has released the weights for Mistral Large 4, also called “Le Chonk,” following the model’s earlier public preview. The company describes it as a 1-trillion-parameter, natively multimodal model trained from scratch on European infrastructure.

According to Mistral, three additional weeks of reinforcement learning improved the model’s coding and defensive cybersecurity performance. The company says Large 4 scored 19.7% on Cognition’s private FrontierCode 2.0 benchmark, led open-weight models on Terminal-Bench 4.0 with a 41% score, and matched the leading open-weight model on the Artificial Analysis Coding Agent Index while using fewer tokens per task. These results are claims reported by Mistral and were not independently verified in the supplied material.

The model offers three reasoning modes: low for simple tasks, medium for everyday work and high for demanding use cases. Mistral says the modes are intended to give developers more control over the balance between performance and inference cost.

For cybersecurity, Mistral reports a 51% score on the Artificial Analysis Cyber Index and an average 96.6% refusal rate for malicious cyber requests. The company says the model is designed to support defensive activities such as vulnerability analysis, reproduction and patching while distinguishing them from offensive requests. The announcement also cites testing or support from organizations including Booz Allen Hamilton, Cisco, Cloudflare, CrowdStrike, Elastic and ENISA.