How much does Voxtral 3 cost?

Mistral prices Voxtral 3 transcription per minute of audio: $0.0035 for Voxtral Mini Transcribe 3 in batch, $0.0040 for Voxtral Small Transcribe 3 and $0.008 for Voxtral Mini Transcribe Realtime 3. The Voxtral Speech 3 text-to-speech model costs $13 per million characters, with the same price for batch and realtime use through one API.

Are the Voxtral 3 models open weights?

The Voxtral 3 announcement does not link any open weights or a license for download. Mistral offers the models on its API with European, US and global endpoints and in the Studio playground. Teams that need audio to stay in their own environment can ask Mistral for self-hosted or private-cloud deployment.

How accurate is Voxtral Small Transcribe 3 compared with Whisper and other APIs?

On Mistral's FLEURS test across 45 languages, Voxtral Small Transcribe 3 posts a 4.0% word error rate with inferred language, against 4.5% for Gemini 3.5 Transcribe, 4.9% for Scribe v2, 5.6% for GPT Transcribe and 14.9% for Whisper large-v3. Microsoft AI Transcribe 2 ties it at 4.0%. These are Mistral's own published numbers.

What does the Voxtral voice agent SDK include?

Mistral's Voxtral voice agent SDK is built on LiveKit. It listens with Voxtral Mini Transcribe Realtime 3, reasons and calls tools with a general model such as Mistral Small, and replies with Voxtral Speech 3. Mistral says each turn completes in under one second, the agent stops when the caller interrupts, and it works over SIP telephony and WebRTC.

Can Voxtral Speech 3 clone a voice?

Voxtral Speech 3 supports voice cloning from 30 or more minutes of audio, according to Mistral. A Pro Voice option adapts a speaker's character, emotional range and style across languages from about 30 minutes of recordings. Prompts can also set emotion, pacing and emphasis directly as instruction variables.

Mistral AI · 2026-10-20 · major

Voxtral 3 — Mistral's new speech-to-text, text-to-speech and voice agent SDK

Voxtral 3 is Mistral's new speech family: Small and 4B Mini transcription models, a sub-500ms realtime model, the Voxtral Speech 3 TTS model and a LiveKit-based voice agent SDK. Transcription starts at $0.0035 per minute.

Mistral AI Voxtral 3 announcement banner

Mistral ships a full speech stack: three transcription models, a TTS model and an SDK that joins them into a voice agent.

Key specs

Fleurs wer (small)	4.0%
Realtime latency	<500ms

Quick facts

Maker	Mistral AI
Models	Small Transcribe 3, Mini Transcribe 3 (4B), Mini Transcribe Realtime 3, Voxtral Speech 3
Transcription languages	50+ (realtime: 35+)
TTS languages	35+ languages, 50+ dialects
Availability	Mistral API (EU, US, global endpoints) + Studio playground
Self-hosting	On request (contact Mistral)

Benchmarks

Keyphrase recall, Earnings 22 (with context)

Voxtral Small Transcribe 3		94.6%
Scribe v2		94.3%
Gemini 3.5 Transcribe		87.3%

source ↗

Pricing

Voxtral Mini Transcribe 3 · batch	$0.0035 / minute
Voxtral Small Transcribe 3	$0.0040 / minute
Voxtral Mini Transcribe Realtime 3	$0.008 / minute
Voxtral Speech 3 (TTS) · batch and realtime	$13 / 1M characters

source ↗

What is it?

Voxtral 3 is a new family of Mistral speech models plus a voice agent SDK. It has two transcription models (Small, and a 4B Mini) for 50+ languages, a realtime transcription model, and Voxtral Speech 3 for text-to-speech in 35+ languages.

How does it work?

The transcription models label who said what (diarization), accept keyphrases or free text to bias toward names and jargon without fine-tuning, and detect the end of a turn from both sound and meaning. The realtime model streams text in under 500ms. The SDK chains realtime listening, a reasoning model with tool calls, and Voxtral Speech 3 replies in one streaming pipeline.

Why does it matter?

Contact centers and assistant builders usually stitch together separate ASR, LLM and TTS vendors. Mistral now sells all three pieces on one API, with diarization and context biasing built in and transcription under half a cent per minute.

Who is it for?

developers building voice agents and transcription pipelines

Frequently asked questions

How much does Voxtral 3 cost?: Mistral prices Voxtral 3 transcription per minute of audio: $0.0035 for Voxtral Mini Transcribe 3 in batch, $0.0040 for Voxtral Small Transcribe 3 and $0.008 for Voxtral Mini Transcribe Realtime 3. The Voxtral Speech 3 text-to-speech model costs $13 per million characters, with the same price for batch and realtime use through one API.
Are the Voxtral 3 models open weights?: The Voxtral 3 announcement does not link any open weights or a license for download. Mistral offers the models on its API with European, US and global endpoints and in the Studio playground. Teams that need audio to stay in their own environment can ask Mistral for self-hosted or private-cloud deployment.
How accurate is Voxtral Small Transcribe 3 compared with Whisper and other APIs?: On Mistral's FLEURS test across 45 languages, Voxtral Small Transcribe 3 posts a 4.0% word error rate with inferred language, against 4.5% for Gemini 3.5 Transcribe, 4.9% for Scribe v2, 5.6% for GPT Transcribe and 14.9% for Whisper large-v3. Microsoft AI Transcribe 2 ties it at 4.0%. These are Mistral's own published numbers.
What does the Voxtral voice agent SDK include?: Mistral's Voxtral voice agent SDK is built on LiveKit. It listens with Voxtral Mini Transcribe Realtime 3, reasons and calls tools with a general model such as Mistral Small, and replies with Voxtral Speech 3. Mistral says each turn completes in under one second, the agent stops when the caller interrupts, and it works over SIP telephony and WebRTC.
Can Voxtral Speech 3 clone a voice?: Voxtral Speech 3 supports voice cloning from 30 or more minutes of audio, according to Mistral. A Pro Voice option adapts a speaker's character, emotional range and style across languages from about 30 minutes of recordings. Prompts can also set emotion, pacing and emphasis directly as instruction variables.

Try it

Model ID voxtral-small-transcribe-3-0 on the Mistral API (console.mistral.ai)