Skip to content

Mistral Small 3.1

An open 24B base model for text and vision tasks

Visit official site

What Mistral Small 3.1 is

The model is described as an open-source base model with 24 billion parameters that supports multilingual and long-context processing for text and vision tasks.

What you can do with it

  • Analyze an image and generate descriptive text
  • Understand and generate text in multiple languages
  • Hold in-depth conversations over long text

Getting started

  • Install the latest vLLM library with pip
  • Load Mistral-Small-3.1-24B-Base-2503 by model name
  • Prepare text and image inputs as needed
  • Encode the inputs into the format the model expects

Model notes

  • Multimodal analysis handles text and visual input
  • A 128k context window handles long text
  • The Tekken tokenizer has a 131k vocabulary
  • The Apache 2.0 license covers commercial and non-commercial use

Pros & cons

✓ What we like

  • Handles text and visual input at the same time
  • Supports dozens of languages

! What to watch out for

  • The listed setup runs through vLLM with a pip install
  • Inputs must be encoded into the format the model expects

FAQ

How large is the context window?

The material states a 128k context window for long text.

Which license applies?

The model uses the Apache 2.0 license for commercial and non-commercial use.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review