Mistral releases weights for Le Chonk model

Summary

Mistral has released the full weights of its one-trillion-parameter open-weight model, Mistral Large 4 ("Le Chonk"). The model is now available for download via Hugging Face, allowing companies and governments to deploy it on their own infrastructure. Mistral claims the model delivers state-of-the-art performance, particularly in agentic coding and cyber defense. The company notes that the model uses 47% fewer tokens per task compared to the Chinese open-weight model GLM-5.3.

Why it matters

Releasing the weights of a one-trillion-parameter model marks a shift in the balance between open and closed AI. European governments and companies gain a sovereign alternative, reducing their dependence on US-based closed labs like OpenAI and Anthropic. This development mitigates data sovereignty risks in cybersecurity. Furthermore, by demonstrating that state-of-the-art performance can be achieved with significantly less computational power and capital, Mistral challenges the pricing and capability narratives of closed providers. For the EU, this strengthens the argument that strategic AI capacity need not be ceded to a single foreign supplier.

Highlights

• The weights are accessible via Hugging Face and the Mistral API, enabling fully on-premises deployment.
• Mistral claims the model leads all open models on the Artificial Analysis Coding Agent Index for agentic coding tasks.
• The model is explicitly positioned for sovereign cyber defense, providing real-time threat detection, analysis, and intervention in code and IT infrastructure.

Critical take

All performance claims in this article—comparability with GLM-5.3, the 47% token efficiency advantage, and leadership on the Coding Agent Index—derive solely from Mistral's own press materials. The article lacks independent benchmarks, third-party expert commentary, license restrictions, hardware requirements, or information on known weaknesses. As a single-source, press-release-style piece from an industry blog, it provides a weak foundation for such a significant news item.