Large Language Models
Oct 29, 2026
Mistral Releases Weights for 1-Trillion-Parameter Model Mistral Large 4
Oct 29, 2026
AI Summary
Mistral has released the weights for its Mistral Large 4 model, a 1-trillion-parameter multimodal AI system. This model is designed to improve coding efficiency and cybersecurity, offering customizable deployment options for organizations seeking sovereign AI solutions.
- Mistral Large 4, also known as Le Chonk, is a 1-trillion-parameter multimodal AI model now available with improved performance in coding and cybersecurity.
- The model supports three reasoning modes to balance performance and cost, making it adaptable for various tasks.
- Mistral Large 4 leads open-weight benchmarks in coding, scoring 19.7% on the FrontierCode 2.0 benchmark and 48% on the Artificial Analysis Coding Agent Index, while using fewer tokens than competitors.
- The model has shown significant improvements in cybersecurity, scoring 51% on the Artificial Analysis Cyber Index, and has a high refusal rate of 96.6% on malicious requests.
- Mistral Large 4 has been validated by cybersecurity vendors and is being tested in real-world environments for its effectiveness in vulnerability detection and patching.
- The model is trained on European infrastructure, emphasizing transparency and control for organizations looking to deploy AI solutions that comply with data residency requirements.
- Mistral is committed to advancing open-weight AI models, enabling developers and enterprises to customize and control their AI deployments effectively.
mistralopen sourcelanguage modelweightsai research