Mistral Large 4
The model uses a granular Mixture-of-Experts design with 52B active parameters, 1.05T total parameters and a 1.6B vision encoder. The API is live through Mistral Studio. Reuters reported that the weights are scheduled for public release on October 27, 2026. Preview pricing of $0.68 per million input tokens and $2.09 per million output tokens is half the standard $1.36/$4.18 rate. Third-party results vary. Artificial Analysis gives it 38 on its Intelligence Index, comparable to GPT-6 Luna (max, 38) and DeepSeek V4.1 Flash (max, 39), which makes it the top model from outside the US and China. The same firm found it costs over 4x more per task than open-weight models of similar intelligence. Vals AI ranks it 32 of 44 on the Vals Index, with its best result 6 of 75 on Harvey's Legal Agent Benchmark. For builders, this is a credible European open-weight option for cyber, legal and finance workloads where self-hosting or data residency matters. Token efficiency and whether the weights ship on schedule will decide if it replaces Chinese open models in production.
