Mistral Large 4 (Le Chonk) is now Open Weight (mistral.ai)

Mistral has officially released Mistral Large 4 (Le Chonk), a 1-trillion-parameter multimodal model that boasts significant advancements in coding, cybersecurity, and reasoning efficiency. Now fully open-weight, this model excels in leading open-weight benchmarks, particularly in agentic coding and defensive security, while providing three reasoning modes to optimize cost and performance. With its European training origins and emphasis on sovereignty, Mistral Large 4 empowers organizations to customize and deploy AI securely on their infrastructure. This release is pivotal for the AI/ML community as it not only sets a new bar for open-weight models, particularly in coding tasks where it has achieved substantial efficiency gains, but it also reinforces cybersecurity capabilities with high accuracy in vulnerability management and patching. For instance, it leads the Artificial Analysis Cyber Index with a 51% score while maintaining a 96.6% refusal rate for malicious requests, showcasing its potential for defensive applications. Mistral’s commitment to open models facilitates greater control for developers and enterprises, allowing for customized solutions and minimizing dependency on external APIs, thus enhancing their AI strategies.