Mistral Voxtral 3 Lands for Voice Agents

Voxtral 3 is less a standalone model drop than Mistral’s push to own the full voice-agent stack.

• –Voxtral Small Transcribe 3 offers a 32K context window, context biasing, diarization, translation, summarization, and function calling. [Model documentation](https://docs.mistral.ai/models/voxtral-small-transcribe-3-0)
• –Mistral claims state-of-the-art multilingual transcription against Whisper, Scribe v2, Gemini 3.5 Transcribe, and other competitors; those claims still warrant independent validation.
• –Voxtral TTS 3 adds zero-shot and professional voice cloning, emotion and pace controls, custom pronunciation, and streaming output. [TTS documentation](https://docs.mistral.ai/models/voxtral-mini-tts-3-0)
• –The biggest developer advantage is composability: transcription, reasoning, tool calls, and speech generation can run as separate stages, making it easier to swap models or deploy privately.
• –EU-hosted API access and on-premises deployment strengthen Mistral’s sovereignty pitch for contact centers and regulated voice workflows.