Mistral Large 2 (24.11), also known as Mistral Large 2.1, is an updated version of Mistral AI's flagship dense language model released in November 2024. It maintains a parameter count of 123 billion, positioned to offer high-tier reasoning, knowledge, and coding capabilities. The model is designed for efficiency, optimized to run on a single NVIDIA H100 node while competing with significantly larger models in its performance class.
This update specifically improves instruction following, system prompt adherence, and function calling accuracy over the original July 2024 release. It features a 128,000-token context window and is engineered for robust performance in retrieval-augmented generation (RAG) and complex agentic workflows. The model exhibits high proficiency in over 80 programming languages and dozens of natural languages, including French, German, Spanish, Chinese, and Japanese.