From Mistral · · 2m
Voxtral 3: Voice Agent SDK & Speech Models in 50+ Languages
Mistral launches Voxtral 3 voice agent SDK with real-time speech models for 50+ languages.
In brief
Mistral launches Voxtral 3 voice agent SDK with real-time speech models for 50+ languages. Mistral releases Voxtral 3, a voice agent SDK and speech models supporting 50+ languages for real-time multi-step conversations. Originally reported by Mistral.
Voxtral 3 Launch
Mistral introduces Voxtral 3: a voice agent SDK and speech models for real-time multi-step conversations across 50+ languages.
Real-Time Speech Capabilities
Voxtral 3 supports accurate transcription across 50+ languages, diarization, and context biasing for improved accuracy.
Natural Speech Generation
Voxtral Mini TTS 3 generates expressive speech with emotion, pace, and emphasis across 35+ languages and 50+ dialects.
Voice Agent SDK
The SDK combines listening, reasoning, tool calling, and speaking into a single real-time streaming pipeline for hands-free interactions.
Sovereign Deployment
Voxtral 3 offers EU-hosted API or self-deployment for data sovereignty, ensuring sensitive conversations stay within your infrastructure.
Multilingual Transcription
Voxtral Small Transcribe 3 achieves state-of-the-art multilingual transcription with diarization and context biasing across 50 languages.
Realtime Transcription
Voxtral Mini Transcribe Realtime 3 delivers high-quality transcriptions at sub-500ms latency across 40+ languages with diarization.
Voice Cloning
Pro Voice adaptation reproduces a speaker’s distinctive character, emotional range, and speaking style across languages using 30 minutes of audio.
Pricing
Voxtral Mini Transcribe 3 costs $0.0035/min, Voxtral Mini TTS 3 costs $13 per 1 million characters.
Performance Benchmarks
Voxtral Small Transcribe 3 outperforms Whisper, Scribe v2, and Gemini 3.5 Transcribe in multilingual transcription accuracy.
Real-World Use Cases
Voxtral handles customer support, appointment booking, and complex tasks like flight bookings and medical consultations.
Diarization
Voxtral models detect who spoke when in conversations, improving accuracy for multi-speaker interactions.
Context Biasing
Key terms and domain-specific vocabulary improve transcription accuracy without requiring model fine-tuning.
Voice Code Bench
Voxtral leads in pronunciation control, multilingual support, and expressiveness across enterprise customer support scenarios.
Deployment Options
Voxtral 3 can be deployed via Mistral's EU-hosted API or self-deployment on your infrastructure for data control.
Next Steps
Mistral invites users to test Voxtral in Studio playgrounds for voice agents, transcription, and speech generation.
What to remember
Mistral releases Voxtral 3, a voice agent SDK and speech models supporting 50+ languages for real-time multi-step conversations.
