Skip to main content
Rase AI Intelligence Newsletter

Creative AI Model or AI service

Voxtral adds speech understanding to downloadable language models

Mistral’s original Voxtral release combines speech recognition with a language model that can answer questions about audio. Its Mini and Small variants have downloadable weights. [1][2][4]

Model access
Original Voxtral Mini 3B and Small 24B have downloadable weights. [1][2][4]
License
Apache License 2.0 for the original Mini and Small weights. [3][4]

What the model does

Mistral released the 3B Mini and 24B Small models on July 15, 2025. The company documented both transcription and interactive tasks such as summarizing an audio clip or answering a question about it. [1][2][4]

Why it matters

A team can use an audio model to answer questions about a recording without relying only on a separate transcript. Comparing that answer with the recording helps catch errors that a fluent summary can hide. [2][4]

Who can use it

Developers working on audio search, transcription or recorded interviews can evaluate the two published variants. The larger checkpoint needs more resources than Mini. [1][2][4]

Key dates

  1. Original Voxtral Mini and Small released [1][2]

Ways to use it

For a recorded presentation, test a transcript and a short answer to a specific question. Check important names or numbers against the original audio before reusing them. [2][4]

Model limitations

This profile covers the original July 2025 checkpoints. Limits and hosted availability vary with the exact Voxtral version and endpoint. [1][2]

Sources and verification

Catalog evidence checked How we verify records

  1. Primary source Mistral model documentation for original Voxtral Mini
  2. Primary source Mistral model documentation for original Voxtral Small
  3. Primary source Original Voxtral Small weights and model license
  4. Primary source mistralai/Voxtral-Mini-3B-2507
  5. Other source Voxtral | Mistral AI
  6. Other source Voxtral

Related profiles

Creative AI

Seedance 2.5 (4K API and Flash)

ByteDance (Volcano Engine / BytePlus)

These tools generate video through developer services. The cited sources describe different variants and routes of access.

Medium confidenceEvidence checked

Creative AI

LongCat-Video-Avatar 1.6

Meituan (LongCat team)

It generates video of an avatar from supported inputs in the documented project.

High confidenceEvidence checked

Creative AI

MiniMax H3-Image

MiniMax

It generates images from prompts through the model architecture described by MiniMax.

High confidenceEvidence checked