Mistral Large 2

Mistral Large 2 is a 123 billion parameter dense language model that Mistral AI released on July 24, 2024, replacing the original Mistral Large as the company’s flagship offering. It handles a 128,000-token context window, scores 84.0% on MMLU, and Mistral trained it with heavy emphasis on multilingual ability, covering languages including French, German, Spanish, Chinese, Japanese, Korean, Arabic, and Hindi alongside English, plus more than 80 programming languages for code generation.

It was a closed, proprietary model available through Mistral’s API and Azure, priced at $2 per million input tokens and $6 per million output tokens, and it also powered Mistral’s Le Chat assistant. Mistral Large 3 has since replaced it as the company’s frontier model, but Mistral Large 2 was notable at the time for closing much of the performance gap with GPT-4 and Claude 3 at a fraction of the parameter count of the largest closed models.