Paris-based Mistral AI has released its latest model, Mistral Large 4, featuring 1 trillion parameters. The model utilizes a mixture of experts architecture, where 49 billion parameters are active per query. The company plans to release the model weights by the end of October, allowing external developers to download and evaluate the system.
The model is priced at $1.36 per million input tokens and $4.18 per million output tokens. This pricing structure positions the model significantly lower than competitors like Claude Opus 5.5 and GPT-6 Astra. Mistral AI markets these capabilities as sovereign AI, designed for organizations that require local data control.
In September, Mistral AI secured a Series D funding round of €3 billion, reaching a valuation exceeding €21 billion. The round was led by Samsung. This capital is intended to support the company's development roadmap, with Mistral Large 4 serving as the first major milestone following the investment.
The model has undergone various performance evaluations, including the AutomationBench test, where it scored 59.9 points. In a blind coding quality assessment by Surge AI, it achieved a score of 3.74 out of 5. Additionally, the model recorded a score of 62 on the DeepSWE 1.1 benchmark, outperforming several Chinese open-weight models.
