N環球 
France's Mistral Launches Voxtral 3: New Voice Model Family and Real-Time Voice Agent SDK
Mistral AI released Voxtral 3 on October 20 — a full voice agent stack with transcription, sub-500ms realtime transcription, TTS in 50+ languages and a LiveKit-based Voice Agent SDK.
Why it matters
French AI company Mistral AI announced on its official blog on October 20 that Voxtral 3 is "available today" — a complete voice agent stack combining listening, reasoning, tool calling and speaking into a single real-time streaming pipeline. It is Europe's most concentrated push yet into the voice-agent race.
What's included
Voxtral Small Transcribe 3 for speech-to-text in 50+ languages, including 24 EU official languages; Voxtral Mini Transcribe Realtime 3 for realtime transcription with sub-500ms latency; Voxtral Speech 3 for text-to-speech in 35+ languages and 50+ dialects; and a Voice Agent SDK built on LiveKit.
Pricing and deployment
Official pricing starts at $0.0035/min for transcription, $0.008/min for realtime transcription, and $13 per million characters for TTS; the voice agent API (with Mistral Small 4 reasoning) costs $0.04/min. Deployment is via Mistral's Europe-hosted API or self-hosting, pitched as "sovereign deployment" with data control.
Performance claims (Mistral's own)
Mistral says Voxtral Small Transcribe 3 reaches a 4.1% word error rate on FLEURS-45, versus 16.0% for Whisper large-v3. These are vendor-reported figures; independent verification is pending.
💬 Comments
No comments yet. Be the first!