Mistral AI has released an update to its open-source language model, upgrading Mistral Small 3.1 to Mistral Small 3.2. The 24-billion parameter model features improvements in instruction following, reducing infinite generation and repetitive output issues, and enhanced robustness in function calling templates. The update also brings a more refined tone and better handling of challenging prompts. Mistral Small 3.2 is available under the Apache 2.0 license and has been integrated into platforms such as LM Studio and Ollama. Additionally, developers have addressed tool calling compatibility for frameworks like GGUF and transformers, including support for date ranges from 2024 to 2028 and experimental FP8 quantization. This release aims to provide a smarter and more reliable model experience while maintaining open access for the AI community.