The latest artificial intelligence model news of the day features Mistral, a company from France, unveiling their new small model, Mistral Small 3.1. This is an upgraded version of Mistral Small 3.0, which was just released last month. While the parameter size of 24B remains the same, its capabilities have been enhanced, allowing it to outperform comparable models like GPT-4o-mini, Claude 3.5 Haiku, and Gemma 3.
Previously, Mistral Small 3.0 scored lower than GPT-4o-mini but excelled in response time, with a speed of 150 tokens per second. With the release of version 3.1, it maintains the same speed while increasing its scoring capabilities, now surpassing all equivalent models.
In addition to speed and capabilities, another highlight of Mistral Small 3.1 is its compatibility with the GeForce RTX 4090 GPU (similar to Google’s Gemma 3) or running on a Mac machine with 32GB RAM, making it ideal for local deployments.
Mistral Small 3.1 is licensed under Apache 2.0 and is now available for download from Hugging Face, as well as being offered on Google Vertex AI and Mistral’s La Plateforme.
TLDR: Mistral introduces their upgraded AI model, Mistral Small 3.1, with enhanced scoring capabilities, speed, and compatibility with advanced GPUs, catering to local deployments.
Leave a Comment