Artificial Analysis says Mistral Large 4 is the most intelligent model from outside the US and China, achieving results comparable to DeepSeek V4.1 Flash (max)
Mistral has released Mistral Large 4, scoring 38 on the Artificial Analysis Intelligence Index; France is back to having …
Context & Ripple Effects
Mistral’s model line has moved from the 123B-parameter Large 2 in 2024 to the Apache-licensed Mistral 3 family in 2025, then to Small 4’s unified reasoning, multimodal and coding design in 2026. Large 4 extends that progression with a 1T-parameter open-weight flagship preview.
Artificial Analysis provides an independent comparison point for Mistral’s performance claim, placing the model alongside DeepSeek V4.1 Flash rather than only against Mistral’s prior releases. Mistral has said the weights are due October 27.
First-order effects
- Mistral gains a third-party intelligence score of 38 and a concrete performance comparison with DeepSeek V4.1 Flash, strengthening its pitch for Large 4 ahead of the planned weight release.
- Developers evaluating open-weight frontier models have an external benchmark for Mistral Large 4, alongside Mistral’s claim that it leads open models developed in the US and Europe.
Second-order effects
- DeepSeek becomes a more explicit competitive reference point for Mistral, pushing both vendors’ evaluations toward comparable model quality rather than regional positioning alone.
- Mistral’s forthcoming weights give enterprises and model hosts a reason to evaluate Large 4 against the earlier Apache-licensed Mistral 3 family when they become available.
Third-order effects
- If high-scoring open-weight models continue to emerge beyond the US and China, access sovereignty becomes less dependent on a small set of closed-model providers.
- Independent benchmark firms gain influence over model procurement as their scores become a common shorthand for comparing releases across regions and vendors.
The trend: Frontier-model competition is broadening into an open-weight race in which third-party benchmarks shape credibility across regional AI ecosystems.