Mistral small 3.1

Introducing Mistral Small 3.1. Multimodal, Apache 2.0, outperforms Gemma 3 and GPT 4o-mini.

Mistral Small 3.1 is a cutting-edge, open-source AI model that excels in its weight class, outperforming competitors like Gemma 3 and GPT-4o Mini. It features multimodal capabilities, multilingual support, and an expanded context window of up to 128k tokens, delivering fast inference speeds of 150 tokens per second. Designed for versatility, it handles tasks such as instruction following, conversational assistance, image understanding, and function calling, making it ideal for both enterprise and consumer applications. Its lightweight design allows it to run efficiently on devices like a single RTX 4090 or a Mac with 32GB RAM. Mistral Small 3.1 is available for download and testing via platforms like Hugging Face, Google Cloud Vertex AI, and Mistral AI’s developer playground, with upcoming availability on NVIDIA NIM and Microsoft Azure AI Foundry.