# Mistral releases Voxtral 3 speech models and voice agent SDK

_Published Tuesday, October 20, 2026 at 12:05 PM EDT · AI · Latest · Tier 2 — Notable_

![Mistral releases Voxtral 3 speech models and voice agent SDK — Primary](https://mistral.ai/cms-media/api/media/file/hero-blog-voxtral.png)

Mistral made Voxtral 3 speech models and a voice agent SDK available through its API, combining transcription, reasoning, tool calls and speech generation for live conversations. The company also offers deployment on customers' own infrastructure through its team.

The SDK uses LiveKit to connect separate models in a streaming pipeline, allowing developers to change the reasoning model, voice or language. Agents can retrieve information and act through business systems during phone or web conversations. Mini batch transcription costs $0.0035 per minute, and speech generation costs $13 per 1M characters. Mistral says the pipeline completes each turn in under one second.

## Sources

- [Mistral AI Blog](https://mistral.ai/news/voxtral-3/)

---
Canonical: https://techandbusiness.org/newswire/meLuRyuB3FV5QJv8u5Tyye
Published: 2026-10-20T16:05:04.950Z
Story chronology: 2026-10-20T12:00:22.000Z
Retrieved: 2026-10-20T19:30:25.694Z
Publisher: Tech & Business (techandbusiness.org)