The whole dayTue 20 Oct 2026
AI Compass
ArchiveAbout

ModelMistral AIConfirmed

Story 2 of 9 in the Tuesday 20 October edition

Mistral AI’s Voxtral 3 runs listening, tools and speech as one pipeline

Voxtral 3 is a family of speech-to-text and text-to-speech models plus a voice agent SDK, on Mistral’s API or arranged for a company’s own machines.

Edited by Ashutosh Shrivastava
21 Oct 2026, 10:22 IST · 2 min read

Abstract artwork with a black field on the left and overlapping blue rectangles, one of them red, on the right
Image: Mistral AIMistral’s image for Voxtral 3

Mistral AI released Voxtral 3 on 20 October 2026: speech-to-text models, text-to-speech models and a voice agent SDK. The transcription models are Voxtral Mini Transcribe 3, Voxtral Mini Transcribe Realtime 3 and Voxtral Small Transcribe 3. The audio docs name the speech models Voxtral TTS 3 and Voxtral TTS Realtime 3. The API is on European, US and global endpoints.

The SDK, built on LiveKit, keeps listening, reasoning and speaking in one stream. Mistral says Voxtral Mini Transcribe Realtime 3 transcribes at sub-500ms and detects the end of a turn. A general model such as Mistral Small then chooses the next step and can call tools. Mistral says the stages do not wait on one another, and a turn finishes in under one second.

On Mistral’s FLEURS comparison for 45 languages, Voxtral Small Transcribe 3 is level with Microsoft AI Transcribe 2 and ahead of Gemini 3.5 Transcribe, Scribe v2 and Whisper large-v3.

Voxtral Small Transcribe 3 matches Microsoft AI Transcribe 2 at 4.0% word error when the language is given and 4.1% when it is detected, ahead of Whisper large-v3 at 14.9% and 16%
FLEURS word error over 45 languages. First value, language given. Second, language detected.Image: Mistral AI

The other chart is diarization, the task of telling speakers apart.

Voxtral Small Transcribe 3 records 18.4% diarization error, just ahead of Voxtral Mini Transcribe 3 at 18.6% and Pyannote at 19%, with Gemini 3.5 Transcribe at 38.4%
Diarization error on Mistral’s public set. Lower is better.Image: Mistral AI

The news post calls the speech model Voxtral Speech 3 and prices it at $13 per million characters. The price table names Mistral Small 4 for the cheaper voice-agent rate, while the pipeline text says a model such as Mistral Small. Realtime transcription is given as more than 35 languages in one place and 40 or more on a product card on the same page.

Key numbers

  • Voxtral Mini Transcribe 3: 4B parameters, $0.0035 per minute in batch
  • Voxtral Small Transcribe 3: $0.0040 per minute
  • Voxtral Mini Transcribe Realtime 3: $0.008 per minute
  • Voice agent API: $0.04 per minute with Mistral Small 4, or $0.08 with another reasoning model

Mistral has not published the weights. Its pricing page says speech models are billed per minute and does not list these rates.

Sources

  1. PrimaryVoxtral 3 announcementmistral.ai · Models, prices, charts and the voice agent pipeline
  2. OfficialAudio overviewdocs.mistral.ai · The five model names used in the docs
  3. OfficialPricingmistral.ai · Speech models are billed per minute
  4. On X@MistralAI on X: Introducing Voxtral 3x.com · Mistral’s launch post, 20 October 2026

The rest of the day

Read the whole day

Spotted an error? Email aiforsuccess.in+compass@gmail.com or tag @ai_for_success on X