Mistral: Mistral Small 3.1 24B

Mistral Small 3.1 24B Instruct is an advanced variant of Mistral Small 3, boasting 24 billion parameters and enhanced multimodal capabilities. It delivers state-of-the-art performance across various tasks, including text-based reasoning, complex vision analysis, programming, mathematical problem-solving, and comprehensive multilingual support in dozens of languages. Equipped with an extensive 128k token context window and optimized for efficient local inference, this model is ideal for diverse applications such as sophisticated conversational agents, robust function calling, in-depth long-document comprehension, and privacy-sensitive deployments. It offers a maximum output of 4K tokens and is available for FREE access, with competitive pricing at $0.03/0.11 per 1M tokens (input/output). The updated version is Mistral Small 3.2.