Mistral opens Large 4 weights for coding and security
Open weights for coding, cybersecurity and enterprise agents
Open-weight release
Mistral has released the weights of Mistral Large 4, codenamed Le Chonk, opening its 1-trillion-parameter flagship for coding, cybersecurity and enterprise agents.
The public preview arrived earlier this month, and Mistral says three weeks of extra reinforcement learning brought more efficient reasoning and lower inference costs.
The model scored 19.7% on Cognition's private FrontierCode 2.0 benchmark and reached 48% on Artificial Analysis's Coding Agent Index, matching GLM-5.3 while using 47% fewer tokens per task on average.
Its Terminal-Bench 4.0 score rose 13 percentage points from preview to 41%, while output tokens fell 30%, and Low, Medium and High reasoning modes let developers tune cost against performance, with High used by default.
On cybersecurity, Mistral Large 4 scored 51% on the Artificial Analysis Cyber Index, and Mistral says its guardrails refuse an average of 96.6% of malicious cyber requests, the highest rate it measured among open models.
The company also acknowledged potential uplift for real attacks in live-traffic monitoring and said it is modifying guardrails, while a specialized cyber variant is being tested with vetted enterprise partners on owned or authorized systems.
The Mistral Ecosystem Open-Weight license is a modified MIT license that permits free use, modification, fine-tuning and commercialization, though Model-as-a-Service businesses need a separate commercial license.
From the source
Mistral has released the weights of Mistral Large 4, codenamed Le Chonk, making its 1-trillion-parameter, natively multimodal flagship available as an open-weight model for coding, cybersecurity and enterprise agents. The public preview arrived earlier this month.
Mistral also reports that three weeks of additional reinforcement learning drove further gains, including more efficient reasoning and lower inference costs.
Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity. 💻 Built to build: Le Chonk ranks #1 on @cognition ’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on… pic.twitter.com/R9wQ6WTPnl
The model scored 19.7% on Cognition’s private FrontierCode 2.0 benchmark. On Artificial Analysis’s Coding Agent Index, it reached 48%, matching GLM-5.3 while using 47% fewer tokens per task on average. Its Terminal-Bench 4.0 score rose 13 percentage points from preview to 41%, while output tokens fell 30%. Low, Medium and High reasoning modes now let developers tune cost against performance, with High used by default.
Cognition's private FrontierCode 2.0 benchmark and reached 48% on Artificial Analysis's Coding Agent Index, matching GLM-5.3 while using 47% fewer tokens per task on average.