Model release · Mistral
Voxtral 3
- Model released
- Review updated
The family includes two speech-to-text models, Voxtral Small Transcribe 3 and the smaller, lower-cost Voxtral Mini Transcribe 3, plus Voxtral Mini TTS 3 for speech synthesis. Small Transcribe 3 supports diarization, context biasing and end-of-turn detection. Mini TTS 3 supports more than 35 languages and 50 dialects, using context or explicit tags to control emotion, pace and emphasis. Pro Voice is rolling out professional voice adaptation using up to 30 minutes of recorded audio.
Availability
Mistral's voice agents API and a LiveKit-based SDK; all three models are on OpenRouter with EU-hosted and zero-data-retention endpoints.