What happened
On October 20, 2026, Mistral AI announced Voxtral 3, a family of speech-to-text and text-to-speech models with a voice agent SDK. It includes Voxtral Small Transcribe 3 and Voxtral Mini Transcribe 3 for transcription in over 50 languages with speaker diarization and context biasing, Voxtral Mini Transcribe Realtime 3 for low-latency streaming transcription, and Voxtral Mini TTS 3 for expressive speech in 35+ languages and 50+ dialects.
The SDK, built on LiveKit, streams speech in, reasoning by a general model such as Mistral Small, tool calls and speech out in one pipeline, and Mistral says each turn completes in under a second. Agents can take and place calls over SIP or run in browsers and apps over WebRTC. Pro Voice adapts a speaker's voice from about 30 minutes of audio.
Access and pricing
Voxtral 3 is available through the Mistral API on European, US and global endpoints, and you can try it in the Studio playgrounds. Mistral also offers self-deployment on your own infrastructure; for on-premises or private-cloud use it asks you to contact its team.
Mistral's listed prices: Voxtral Mini Transcribe 3 (batch) $0.0035 per minute, Voxtral Small Transcribe 3 $0.004 per minute, Voxtral Mini Transcribe Realtime 3 $0.008 per minute, Voxtral Mini TTS 3 $13 per million characters, and the voice agent API $0.04 per minute with Mistral Small 4 or $0.08 per minute with a larger model.
What to check
Mistral's charts show its transcription models ahead of Whisper, Scribe v2 and Gemini 3.5 Transcribe and its TTS cheaper than several rivals; these are the company's own results. The page also gives the realtime model's language coverage as both over 35 and 40+ languages, so check the documentation for your language before building.
Go to the source
Sources & further reading
Based on the official documentation below. Features and access can change; check the provider before making a decision.
Was this page useful?
Help us make our coverage more useful. This is feedback on the page, not a rating of the tool.