Mistral released Voxtral 3, a voice agent SDK and speech models. It supports real-time speech in over 50 languages. The system combines accurate transcription, emotional text-to-speech, and deep tool use. It is available via Mistral’s EU-hosted API or self-deployment. Voxtral Small Transcribe 3 offers state-of-the-art multilingual transcription with diarization and context biasing. Voxtral Mini TTS 3 supports 35+ languages and 50+ dialects with emotional adaptability. The SDK integrates listening, reasoning, tool calling, and speaking into one streaming pipeline. Pro Voice adaptation reproduces a speaker’s voice from 30 minutes of audio. Realtime transcription achieves sub-500ms latency in over 35 languages. Pricing starts at $0.0035 per minute for transcription and $13 per 1M characters for TTS.
What changed
Building end-to-end voice agents required separate components for listening, reasoning, speaking, and tool use.
What it unlocks
Creating real-time, multilingual voice agents that transcribe, reason, speak with emotion, and use tools in one pipeline.
- 50+ languages supported
- 35+ languages for TTS
- 50+ dialects for TTS
- $0.0035 per minute for Mini Transcribe
What you need to act on it
- Mistral EU-hosted API access or self-deployment capability
Sources