Model Family
Pixtral
Mistral's open-weights vision-language model â Pixtral 12B and Pixtral Large.
Definition
Pixtral is Mistral's multimodal model family, capable of ingesting variable-resolution images alongside text. Pixtral Large (124B) is competitive with closed-source VLMs while remaining openly downloadable, making it popular for self-hosted document and image understanding.
Common use cases
- Document AI
- OCR
- Visual Q&A