Voxtral Small Transcribe 3
Voxtral Small Transcribe 3 is the larger transcription model in Mistral's October 2026 audio preview. It supports transcription and speaker diarization. Mistral also describes speech translation, questions about audio, summaries and function calling.
Preview
Distinguish the model from its endpoint
The model description lists speech understanding features, while its feature panel specifically links transcription and timestamp operations. An application should confirm the request format for the intended operation. A model capability list alone does not establish that every capability uses the same endpoint.Voxtral Small Transcribe 3 model page
Evaluate the audio workflow
Practical guidance. Use Small when the decision depends on meaning as well as the words spoken. For meeting records, inspect whether a summary preserves who agreed to each action. Compare these errors with Mini on the same recordings before selecting a model for every request.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Preview |
| Launch date | |
| Access | API |
| Model identifiers | voxtral-small-transcribe-3-0 |
| Context window | 32,000 tokens The model page advertises 32k context. This does not specify a maximum recording duration. Voxtral Small Transcribe 3 model page |
| Initial release stage | Public preview.Voxtral Small Transcribe 3 model page |
| API identifier | voxtral-small-transcribe-3-0Voxtral Small Transcribe 3 model page |
| Access | Premier API service; public weights not confirmedVoxtral Small Transcribe 3 model page |
Related models
- Voxtral Small 2507
Mistral AI · Voxtral
An Apache-licensed 24B audio-understanding model that answers questions about recordings and requests tools from spoken instructions.
Launch date API · Downloadable - Voxtral Mini Transcribe 3
Mistral AI · Voxtral
A preview speech-to-text model for recorded audio with timestamps and speaker diarization.
Launch date PreviewAPI