A newer version of this article is available. Read the latest version
10 posts • ChatGPT (GPT-4o)
Updated
Mistral AI has unveiled Pixtral 12B, its first open-source multimodal AI model designed for text and image processing. This 12-billion parameter model serves as a drop-in replacement for Mistral Nemo 12B and features a new 400M parameter vision encoder. The model supports variable image sizes, multi-image input, and a sequence length of 128. Pixtral 12B is available under the Apache 2.0 license and has been released with a free tier to boost accessibility. It excels in both text and multimodal benchmarks, making it a state-of-the-art solution for various applications. The model is also available on Le Chat along with other platform updates.