As of: 2026-10-25 23:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/voxtral-small-transcribe-3/

# Voxtral Small Transcribe 3

Mistral AI, Voxtral family. Status: preview. Type: speech and audio. Released: 2026-10-20.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Mistral API | `voxtral-small-transcribe-3-0` | https://api.mistral.ai/v1/audio/transcriptions | https://docs.mistral.ai/models/voxtral-small-transcribe-3-0 |
| Mistral Studio docs (speech-to-text) |  |  | https://docs.mistral.ai/studio/audio/speech-to-text/overview |

## Pricing

- Per minute: $0.004
- Unit as published: USD per minute of audio (Mistral API)
- Source: https://docs.mistral.ai/models/voxtral-small-transcribe-3-0
- Last checked: 2026-10-25

## Specs

- Input: audio
- Output: text
- Context window: 32,000 tokens
- Max output: not published
- Open weights: no
- Licence: proprietary

## Where it stands

- No newer speech and audio model in the registry yet.
- Before it from Mistral AI: Voxtral TTS (2026-03-23) https://postcutoff.com/m/voxtral-tts/
- Closed weights; access is hosted only.

## What stands out

- Transcription in 50+ languages incl. all 24 official EU languages: Mistral's highest-accuracy transcription model; ~4.0-4.1% macro-average WER on FLEURS (45 languages), which Mistral says beats Whisper large-v3, ElevenLabs Scribe v2 and Gemini 3.5 Transcribe (vendor-reported). Source: https://mistral.ai/news/voxtral-3
- Speaker diarization: Built-in 'who spoke when'; 18.4% macro-average cpWER on Mistral's diarization chart vs 19.0% for the best competitor shown (pyannote). Source: https://mistral.ai/news/voxtral-3
- Context biasing without fine-tuning: Key terms and free-form context text can be passed with the request to improve recognition of names and jargon. Source: https://mistral.ai/news/voxtral-3
- #3 on Artificial Analysis AA-WER (2.1%) (found after launch): Third on the Artificial Analysis non-streaming speech-to-text leaderboard (AA-WER 2.1%, $4.00 per 1,000 min), behind Fun-Realtime-ASR-preview (1.7%) and MAI-Transcribe-2 (2.0%), read 2026-10-25. Source: https://artificialanalysis.ai/speech-to-text

## Notes

Artificial Analysis speech-to-text (https://artificialanalysis.ai/speech-to-text), read 2026-10-25: rank #3, AA-WER 2.1%, $4.00 per 1,000 minutes. Docs label it v3.0, 'Public Preview' and 'Premier' (API-only; no open weights or Hugging Face release found, license page not read). Model card also says it supports translation, speech Q&A, summarization and function calling; context 32k. Exact language list and maximum audio length not stated on the card. Siblings: voxtral-mini-transcribe-3 (4B, $0.0035/min) and Voxtral Mini Transcribe Realtime 3 (see voxtral-mini-transcribe-3). Successor of Voxtral Transcribe 2 (voxtral-transcribe-2) and the July 2025 Voxtral Small (voxtral-small).

Mistral's top-accuracy batch transcription model in the Voxtral 3 line (Oct 2026), EU-hosted API.

```bash
curl https://api.mistral.ai/v1/audio/transcriptions -H "Authorization: Bearer $MISTRAL_API_KEY" \
  -F model=voxtral-small-transcribe-3-0 -F file=@audio.mp3
```

Sources: https://mistral.ai/news/voxtral-3 · https://docs.mistral.ai/models/voxtral-small-transcribe-3-0 · https://docs.mistral.ai/models/overview · https://artificialanalysis.ai/speech-to-text

Model page: https://postcutoff.com/m/voxtral-small-transcribe-3/