Mistral Small 3.2

The official Mistral AI model card on Hugging Face documents Mistral-Small-3.2-24B-Instruct-2506 as a minor update to Mistral-Small-3.1-24B-Instruct-2503, with targeted improvements in three areas: instruction following, reduction of infinite/repetitive generations, and a more robust function calling template. The card The card confirms multimodal input (image + text), supports function calling and structured output, and recommends deployment via vLLM (>=0.9.1, which pulls in mistral_common >=1.6.2) or transformers. The page also notes recommended inference settings (low temperature such as 0.15) and system prompt customization, and