Mistral's Voxtral 3 speech AI models for transcription, text-to-speech with voice cloning, and voice agents
Mistral homepage
1 / 3Overview
What is Mistral Voxtral 3?
Mistral's speech models for transcription, TTS and agents
Voxtral 3 is Mistral's family of speech models for transcription, text-to-speech and voice agents. It includes batch speech-to-text models with speaker diarization and context biasing across 50+ languages, a realtime transcription model with sub-500ms latency, and a text-to-speech model with control over emotion, pacing and accent in 35+ languages and 50+ dialects, plus voice cloning from about 30 minutes of audio.
A voice agent SDK built on LiveKit lets developers create agents that listen, reason, call tools and speak, with function calling, web search, SIP telephony and WebRTC support. Everything is available through the Mistral API (EU, US or global endpoints) or self-hosted and on-premises, with open weights under Apache 2.0.
Pricing is usage-based: batch transcription from $0.0035 per minute, realtime transcription $0.008 per minute, TTS $13 per million characters, and voice agent API from $0.04 per minute. A free Studio tier lets you test transcription, speech generation, voice cloning and voice agents before committing.
Platforms and languages
Mistral Voxtral 3 Availability
Platforms
Languages
Capabilities
Mistral Voxtral 3 Key Features
Batch speech-to-text
Transcribe audio in 50+ languages with speaker diarization and context biasing for better accuracy.
Realtime transcription
Stream live transcription with sub-500ms latency for calls, meetings and live events.
Text-to-speech with voice control
Generate natural speech with control over emotion, pacing and accent, 35+ languages and 50+ dialects.
Voice cloning
Clone a voice from roughly 30 minutes of audio and use it across TTS workloads.
Voice agent SDK
Build voice agents that listen, reason, call tools and speak, with function calling, web search and SIP telephony, via API or self-hosted/on-prem.
Best for
Who uses Mistral Voxtral 3?
Developers and teams building voice products
Add transcription, multilingual TTS or full voice agents to apps, call centers and customer support
Plans and access
Mistral Voxtral 3 Pricing
Freemium
Usage-based API pricing from $0.0035/min transcription; TTS $13 per 1M chars; free Studio tier available.
Common questions
Mistral Voxtral 3 FAQs
Is there a free plan to try Voxtral 3?
Yes. Mistral Studio has a free tier where you can test voice agents, audio transcription, speech generation and voice cloning before paying.
How is Voxtral 3 priced?
Pricing is usage-based through the Mistral API: batch transcription from $0.0035 per minute, realtime transcription $0.008 per minute, TTS $13 per million characters, and voice agent API from $0.04 per minute.
Can I self-host Voxtral 3?
Yes. Voxtral models ship with open weights under Apache 2.0, and on-premises deployments of Voxtral 3 are available through Mistral's sales team.
Which languages are supported?
Transcription covers 50+ languages and TTS covers 35+ languages with 50+ dialects, including English, French, German, Spanish, Italian, Dutch, Polish, Russian, Ukrainian, Chinese and Arabic.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
