Mistral Large 4, codenamed 'le Chonk,' enters public preview on Mistral Studio. The 1 trillion parameter natively multimodal model features 49 billion active parameters and demonstrates state-of-the-art performance across coding, agentic workflows, and multimodal understanding. Weights are scheduled for release by month's end.
Until then, Mistral is red-teaming the model with cybersecurity leaders and vetted partners. Built on 3,800 NVIDIA Grace Blackwell GPUs in European datacenters, the model supports 160+ languages, including all official EU languages. ML4 achieves top-tier cybersecurity capabilities, ranking among the top five globally on the Artificial Analysis Cyber Index and solving 93% of Cybench challenges.
It outperforms other open-weight models outside China, offering significant advantages for enterprise workloads in finance, law, and cybersecurity. The model enables self-deployment under organizational policies, a critical feature for vulnerability research and incident response where provider-level refusals can block legitimate security work. Mistral Forge provides the same training and customization environment offered to customers.
The model will be available across multiple regions worldwide, including a European deployment operated independently under European law.
Source: Hacker News · Summarized by HeadlinesBriefing