Mistral AI releases Voxtral 3 speech models and voice agent SDK
Tech

Mistral AI releases Voxtral 3 speech models and voice agent SDK

TechNews Editorial
TechNews EditorialOct 20, 2026 · 1 min read
Share

Why it matters

According to Mistral, organizations can build voice agents that complete tasks during calls and choose where audio is processed.

The facts

  • Mistral AI released Voxtral 3 speech models and an SDK for voice agents that can use tools during live calls.
  • Mistral says transcription covers more than 50 languages, while speech generation supports more than 35 languages.
  • Voxtral 3 is available now through Mistral’s EU-hosted API or for self-deployment.

Advertisement

Mistral AI has released Voxtral 3, a family of speech models and a voice agent SDK. The models turn speech into text and text into speech. The SDK connects those steps with reasoning and tool calls so an agent can handle tasks during a live conversation.

Voxtral Small Transcribe 3 supports more than 50 languages, according to Mistral. It can identify speakers, use supplied terms to improve transcription of specialized vocabulary, and detect when someone has finished speaking. Mistral also offers a 4-billion-parameter Mini transcription model priced at $0.0035 per minute.

For live use, Mistral says Voxtral Realtime Transcribe 3 works in more than 35 languages with latency below 500 milliseconds. It supports speaker identification and supplied vocabulary. Its end-of-turn detection uses sound and meaning to decide when to respond.

Read nextNvidia releases open models for enterprise agents and local PCs

Speech generation adapts to context

Voxtral Mini TTS 3 generates speech in more than 35 languages and 50 dialects, Mistral says. It can adjust emotion, pace and emphasis based on the conversation or explicit instructions. It is available for batch and real-time use. Mistral lists speech generation at $13 per 1 million characters.

The company also offers Pro Voice adaptation for professional uses. Mistral says it can reproduce a speaker’s voice characteristics and speaking style across languages using 30 minutes of audio.

A parent carrying a toddler pauses outdoors to speak into a phone while the child reaches toward their face.
Illustration: AI & Tech News

The SDK connects speech with tools

The new SDK, built on LiveKit, streams transcription, reasoning and speech generation together. Mistral says agents built with it can retrieve information and use connected systems to check a bill or book an appointment during a phone or web conversation. The company says an agent can respond in under one second per turn and stop speaking when a caller interrupts.

Voxtral 3 is available now through Mistral’s EU-hosted API or for deployment on an organization’s own infrastructure. Mistral says organizations can choose where audio is processed and whether conversations are retained.

Source: Mistral AI

Newsletter

Get the best AI & tech news daily

A concise daily digest. Unsubscribe anytime.

We use your email only to send this newsletter.

Keep reading