Mistral has released Mistral Large 4, a 1 trillion-parameter open-weight model nicknamed Le Chonk, positioning it as the most capable open-weight model developed outside China and very close to leading proprietary systems.
The model is available in preview now, with a final version by month's end. Guillaume Lample, Mistral's cofounder and chief scientist, told WIRED that the lab is targeting domains where larger US-based labs do not focus, emphasizing customization and domain-specific performance over raw benchmark chasing. The company sees its advantage in serving specialized verticals rather than competing head-to-head on general-purpose benchmarks.
Running open-weight models is considerably cheaper for businesses since they only pay for compute consumed. By narrowing the performance gap with proprietary models and providing full access to weights, Mistral is making a direct economic argument for enterprises to switch. The cost savings are significant: businesses can fine-tune and deploy without per-token API charges, and they retain full control over their data and model behavior.
The release arrives amid growing transatlantic tension over AI access. In June, the Trump administration temporarily restricted distribution of models from OpenAI and Anthropic over cybersecurity concerns. The reminder that the US government could revoke access to frontier AI created an opening for Mistral as a Europe-based alternative with no such gatekeeping.