Mistral AI has introduced Mistral Large 4, a 1-trillion-parameter model internally nicknamed Le Chonk. The model is currently available in preview, with a final release scheduled for the end of the month. It is designed for general-purpose tasks but features specific optimizations for coding, cyberdefense, manufacturing, finance, and electrical engineering.
The company claims Le Chonk is the most capable open-weight model developed outside of China and performs closely to leading proprietary alternatives. Unlike some models that rely on distillation, Mistral states that Le Chonk was trained from scratch. The company positions the model as a reliable alternative for businesses concerned about the potential for sudden access restrictions to US-based proprietary AI.
Mistral operates with fewer compute resources and less capital than major US labs like OpenAI and Anthropic. The company generates revenue through pay-as-you-go cloud fees and engineering support for model tuning. In September, Mistral secured a funding round of $3.3 billion, reaching a valuation of $24 billion, which marks the largest raise by a European tech company to date.
The company reports that its earnings have increased 20-fold over the past year. Mistral emphasizes the importance of model ownership, arguing that businesses using closed-source models face risks if access is revoked by providers or government regulations. This strategy aims to appeal to international clients who prioritize technological sovereignty and consistent access to AI infrastructure.
