← All model drops
Live drop
Voxtral 3
Mistral AIDropped Oct 20, 2026
End-to-end voice agent stack with 50-language transcription, expressive TTS and LiveKit SDK for sub-second streaming turns.
About this release
Live drop
Mistral AIDropped Oct 20, 2026
End-to-end voice agent stack with 50-language transcription, expressive TTS and LiveKit SDK for sub-second streaming turns.
About this release
Voxtral 3 is Mistral AI's family of speech-to-text (Small Transcribe 3, Mini Transcribe 3 and realtime variant) and text-to-speech (Speech 3 / Mini TTS 3) models plus a voice-agent SDK. The models handle transcription and diarization across 50+ languages, expressive speech in 35+ languages with emotion/pace control, and integrate with a reasoning LLM and tools in one streaming pipeline. Available on Mistral API (EU/US/global) or self-hosted.
Read the official Mistral AI launch pageFirst builds incoming — this lane fills in automatically as people start building with Voxtral 3. Know one? Submit the URL.