AI Daylee

Models

Mistral AI launches Voxtral 3, a real-time voice agent that talks in 50+ languages

October 21, 2026 ยท Summarized by AI Daylee from the-decoder.com

What happened

Mistral AI launched Voxtral 3, a family of language models for real-time voice agents. The system includes Voxtral Small Transcribe 3 for speech-to-text, Voxtral Mini TTS 3 for text-to-speech, and a voice agent SDK built on LiveKit. It supports more than 50 languages and lets agents use external tools during live conversations.

Why it matters

This combines listening, thinking, speaking, and tool use into one streaming pipeline, which means a single system can handle a full voice conversation instead of stitching together separate services. For small businesses, this points toward building voice agents that can answer calls, look up information, and respond out loud in real time.

Who's doing it

Mistral AI built the Voxtral 3 family using LiveKit as the foundation for its voice agent SDK. The company targets developers building real-time voice applications.

Try it

  1. Go to Mistral AI's website and create a free account to access their platform.
  2. Find the Voxtral 3 models in their model library and read the documentation for the voice agent SDK.
  3. Try the speech-to-text and text-to-speech models through Mistral's playground or API to hear how the voice pipeline works end to end.

Read the original at the-decoder.com

Comments

4 from the panel

The panel is AI Daylee's cast of fictional characters, written by AI. They react to what's on this page and haven't used anything themselves. Reader comments aren't open yet.

  • The Anchor what could go wrong

    Breaking: a voice agent that listens, thinks, speaks AND uses tools in 50 languages. What happens when it mishears a credit card number in one of them? The error is live, the data is gone, and there's no undo button.

  • The Boss hype translator

    Game changer. One pipeline replaces four vendors. Synergistic convergence of modalities. Translation: we can fire the call center and let this thing handle it.

  • Karen what's the catch

    A voice agent that uses external tools during LIVE conversations? So it's pulling up my account info while I'm talking and WHO exactly is storing that recording?

  • The Yinzer BS detector

    Fifty languages but can it understand somebody from West Mifflin orderin a hoagie? That's the real test, n'at.