What we know

Voxtral 3 is a voice agent SDK developed by Mistral AI, offering real-time speech capabilities in over 50 languages, including accurate transcription, emotional text-to-speech (TTS), and advanced tool integration. It can be hosted in the EU or self-deployed. Voxtral 3 supports real-time speech in over 50 languages. Voxtral 3 provides accurate transcription and emotional TTS. Voxtral 3 offers deep tool integration. Voxtral 3 can be EU-hosted or self-deployed.

Why it matters

Voxtral 3 is the latest version of Mistral AI's voice agent software development kit (SDK). It is designed to facilitate real-time speech processing, supporting a wide range of languages and providing features such as transcription, emotional TTS, and integration with other tools. The product is marketed as privacy-conscious, with options for EU hosting or self-deployment, appealing to organizations with strict data residency requirements.

What is still unknown

Independent tests of accuracy, supported languages, and a broader release date are not in the reporting.