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 Large 4 is now the top-ranked AI model developed outside of the US and China, a distinction previously held by South Korea and the UAE.

Mistral has released Mistral Large 4, a new large language model that has achieved a score of 38 on the Artificial Analysis Intelligence Index. This places it among the top-performing models globally, comparable to OpenAI's GPT-6 Luna (max, 38) and DeepSeek V4.1 Flash (max, 39). The model also demonstrates strong capabilities in cyber defense, scoring 50 on the Artificial Analysis Cyber Index, on par with GLM-5.3-Flash. Mistral Large 4 is available in a Research Public Preview, with plans to release its weights at the end of October. Key features include a 512k token context window and multimodal capabilities for text and image input. Despite its advanced intelligence, its cost per task is higher than some competing open-weights models.

Mistral's new Mistral Large 4 model significantly elevates the competitive landscape for AI development outside of the US and China, achieving intelligence benchmarks comparable to leading global players. Its strong performance on both general intelligence and cyber defense indices indicates a maturation of European AI capabilities. The model's open-weights release planned for October will further democratize access to advanced AI, potentially fueling innovation among developers worldwide.

While Mistral Large 4 offers state-of-the-art performance, its pricing model presents a significant cost consideration for widespread adoption. The model's cost per task, even with introductory discounts, is substantially higher than comparable open-weights alternatives, which could influence its immediate uptake by cost-sensitive organizations. This highlights a persistent trade-off between cutting-edge performance and affordability in the current AI market.