Mistral unveils Large 4 AI model, claims edge over Chinese rivals

Technical Highlights

Mistral Large 4 (ML4) carries 1 trillion parameters, with 49 billion active at any given moment. The model is a Mixture-of-Experts system, which sends each request to a small subset of its internal 'experts' rather than firing up the whole network.

Pricing and Availability

ML4 is currently in a limited preview, with full access restricted to select developers and expert users. The model's weights will be released publicly on October 27, 2026. API pricing is also available, with input costs $1.36 per million tokens and output costs $4.18 per million tokens.

Market Outlook

Mistral is targeting ML4 at specific workloads, including cybersecurity, finance, coding, manufacturing, and visual grounding. The model supports more than 160 languages, covering every official language of the European Union.

Regulatory Outlook

The benchmark claims, read carefully According to Mistral, ML4 scored 38 points on the Artificial Analysis Intelligence Index. The company says that is the highest score for any non-Chinese open-weight model. MiMo-V2.6-Pro, one Chinese open model, scored 46 on the same index.

Conclusion

Mistral's ML4 is a significant development in the field of artificial intelligence, and its claims of outperforming some Chinese rivals will be closely watched. As the model's weights are released publicly, it will be interesting to see how it performs in real-world applications.