French AI innovator Mistral AI has released Mistral Small 3.1, a new 24-billion-parameter model designed to perform competitively across text, vision, and multilingual benchmarks.
A key feature of Mistral Small 3.1 is its ability to operate locally without requiring expensive cloud infrastructure. The model runs efficiently on a single RTX 4090 GPU or a Mac with at least 32 GB of RAM when quantized, potentially making it more accessible for startups, developers, and enterprises. Industries ranging from document analysis to medical diagnostics and object recognition stand to benefit from this innovation.
Benchmark breakthroughs
This new AI model is intended as a competitor to Google’s Gemma 3 and OpenAI’s GPT-4o mini, with an expanded 128K token context window and versatile multimodal capabilities, and Mistral Small 3.1 has outscored both in some benchmarks.
Mistral Small 3.1 has demonstrated strong results in various benchmark tests, including text generation, reasoning challenges such as MATH, general knowledge assessments like MMLU, and question-answering tasks (GPQA). The model’s efficiency suggests high performance can be achieved in a more compact framework, contributing to ongoing discussions about the necessity of large-scale AI models.
A key feature of Mistral Small 3.1 is its ability to operate locally without requiring expensive cloud infrastructure. The model runs efficiently on a single RTX 4090 GPU or a Mac with at least 32 GB of RAM when quantized, potentially making it more accessible for startups, developers, and enterprises. Industries ranging from document analysis to medical diagnostics and object recognition stand to benefit from this innovation.
Open-source advantage and versatility
Mistral Small 3.1 has been released under the Apache 2.0 license, allowing developers to modify and deploy the model without licensing restrictions. The 128K token context window supports long-form reasoning and detailed document processing, and the model’s ability to handle both text and image inputs broaden its applications beyond text-based tasks.
Availability and integration
Mistral Small 3.1 is now available for download on the Hugging Face website, offering both the Base and Instruct versions. For enterprise deployments requiring private, optimized inference infrastructure, Mistral AI invites direct contact to tailor solutions. Additionally, users can experiment with the model via API on Mistral AI’s developer playground: La Plateforme. Additionally, integrations with Google Cloud Vertex AI, NVIDIA NIM, and Microsoft Azure AI Foundry are expected to roll out soon.
Mistral Small 3.1 adds to the growing landscape of open-source AI models, providing an alternative to proprietary systems from major technology companies. Its performance and deployment options contribute to ongoing discussions about accessibility, efficiency, and the trade-offs between open and closed AI ecosystems.
Explore our list of top AI companies dominating the AI landscape and stay up to date with the latest advancements in the field.