France-based Mistral, an AI startup, released weights for Mistral Large 4, a trillion-parameter model, on October 29.

The weights are available on Hugging Face for self-hosting, and the model is also accessible through Mistral’s API.

The company is targeting enterprises and governments that want more control over their data and infrastructure, especially for cybersecurity.

Its sovereign AI pitch centers on owning AI infrastructure and data.

Mistral says the model uses 47% fewer tokens per task than GLM-5.3.

It also says the model outperforms other open models on Artificial Analysis’ Coding Agent Index.

Artificial Analysis tracks coding-agent performance, but the reviewed material did not independently confirm that ranking.

The European Commission distinguishes obligations under the European Union AI Act for general-purpose AI model providers from those for deployers.

The reviewed material did not establish any automatic compliance relief for organizations that self-host the model.

🔗 Source: TechCrunch

Recent Mistral developments