Mistral launches Voxtral 3 for transcription and speech
A family covering batch and realtime transcription, text to speech and a voice-agent API. Realtime latency is under 500 ms, with 50+ transcription languages.
Why it matters: Published per-minute prices make it easy to compare with established speech APIs.
- Transcription price
- $0.0035 to $0.008 per minute
- TTS price
- $13 per 1M characters
- Voice agent
- $0.04 to $0.08 per minute
- Parameters
- 4B for Mini Transcribe 3
- License
- Not stated
| Benchmark | Score | Type | Measured or published by |
|---|---|---|---|
| Diarization cpWER (Small Transcribe 3) | 18.4% | Self-reported | Mistral |
Sources: mistral.aimistral.ai