Pixtral Large

Mistral AI · 1 year ago

Mistral AI released Pixtral Large, a 124-billion parameter open-weights multimodal model that achieves the highest scores on MathVista (69.4%), DocVQA, and ChartQA benchmarks while maintaining text understanding performance comparable to Mistral Large 2. The model features a 128K context window capable of processing at least 30 high-resolution images and is available under research and commercial licenses. The model is now accessible via Mistral's API, HuggingFace, and Le Chat, alongside an updated Mistral Large 24.11 text model optimized for enterprise workflows including document understanding and task automation.