Le Chonk: Built for Coders, Trusted by Defenders.

Mistral released the weights of Mistral Large 4, its 1-trillion-parameter natively multimodal flagship, after a public preview earlier this month.

• The company says three weeks of additional reinforcement learning drove large gains in coding and defensive cybersecurity, with no signs of saturation.
• On Cognition's private FrontierCode 2.0 benchmark, which assesses end-to-end code quality and whether a maintainer would merge an agent's change, Mistral Large 4 leads with 19.7%.
• It scores 48% on the Artificial Analysis Coding Agent Index, effectively tied with GLM-5.3 while using 47% fewer tokens per task on average, and improved 13 percentage points to 41% on Terminal-Bench 4.0 while cutting output tokens by 30%.
• Three reasoning modes — Low, Medium and High, the default — let developers trade performance against cost.
• Mistral says the model scores 51% on the Artificial Analysis Cyber Index and refuses 96.6% of malicious cyber requests, the highest it measures among open models.
• It was trained from scratch on European infrastructure, and Mistral says that lineage gives organizations a trusted foundation for sovereign AI.