FOSSWire
Search/
AI Frontier

Mistral Inference v1.4.0 adds support for Pixtral models

Brief written by the FOSSWire AI editor from Mistral AI How we write

1 min read 33 reads

Key Takeaway

The mistral-inference library now supports the Pixtral-12B-2409 multimodal model, allowing for image and text processing.

Summary

Mistral AI has released version 1.4.0 of the mistral-inference repository. This update introduces support for the Pixtral-12B-2409 model, enabling users to run multimodal inference via Python or the command line. The update allows the model to process text prompts alongside one or more images.

Technical details of Pixtral support

  • Supports the Pixtral-12B-2409 model via the mistral_inference package.
  • Enables multimodal input using ImageURLChunk and TextChunk within the ChatCompletionRequest protocol.
  • Includes support for the tekken tokenizer.
  • Provides CLI capabilities via mistral-chat for interactive image and text prompting.

Original reference: github.com →