Mistral has released Voxtral 3, a family of speech models and a software development kit for voice agents that can listen, reason, call tools, and respond in one real-time pipeline. The system is intended for contact centers, service desks, virtual assistants, and hands-free applications.
Voxtral Small Transcribe 3 covers more than 50 languages and includes speaker identification, phrase-based accuracy hints, and end-of-turn detection. Voxtral Mini TTS 3 generates speech in more than 35 languages and 50 dialects, adjusting pace, emphasis, and emotion to the conversation.
The accompanying SDK connects those speech components to reasoning models and business tools, allowing an agent to retrieve information or complete multi-step tasks during a phone or web conversation. Mistral also offers Pro Voice adaptation, which it says can reproduce a speaker’s style across languages from 30 minutes of audio.
Organizations can use Mistral’s EU-hosted API or deploy the stack on their own infrastructure. The self-hosted option gives them control over where audio is processed and whether conversations are retained.