AI

Mistral Releases Le Chonk Weights for Agentic Coding and Cybersecurity

Thursday, October 29, 2026Read Original

Details

  • Mistral AI released the weights for Mistral Large 4, codenamed Le Chonk, on October 29, 2026.
  • The model targets agentic coding, cybersecurity, and enterprise workloads, allowing organizations to download and run the weights on their own infrastructure.
  • Mistral says Le Chonk ranks first among open models on Cognition’s private FrontierCode 2.0 benchmark, scoring 19.7%.
  • FrontierCode evaluates end-to-end coding quality and whether a repository maintainer would accept an agent’s proposed change.
  • The model is a roughly one-trillion-parameter, natively multimodal system designed for complex software and reasoning tasks.
  • Mistral reported that the model uses 47% fewer tokens per task on average than comparable open models on its coding evaluation.
  • Le Chonk’s release follows a preview announced earlier in October and expands access from selected testers to developers and organizations.
  • The weights are available for download, while Mistral also supports access through its API and partner platforms.
  • Mistral trained the model on 4,000 Nvidia Grace Blackwell GPUs over two months, according to company disclosures.

Impact

The release strengthens Europe’s position in the open-weight model race by pairing downloadable weights with performance claims aimed at practical software engineering and cybersecurity. It also pressures rival open models, including GLM-series systems, on coding-agent efficiency and repository-level task completion. Self-hosting could widen adoption among enterprises that cannot send sensitive code or security data to external APIs, although the model’s infrastructure requirements and licensing terms may limit deployment to well-resourced users.

Rift Dispatch