Is Voxtral 3 state of the art?Pending benchmarks.Announced Oct 20, 2026; no independent results yet.
Mistral's new speech-to-text and text-to-speech models, with a voice agent SDK built on LiveKit and Pro Voice cloning from up to 30 minutes of audio. The lab says the TTS model covers 35+ languages and 50+ dialects. Available through the Mistral API; open weights not stated.
The details so far
- Lab
- Mistral
- Announced
- Oct 20, 2026
- API
- Live
- Open weights
- Unknown
- Models
- Voxtral Small Transcribe 3, Voxtral Mini Transcribe 3, Voxtral Mini Transcribe Realtime 3 and Voxtral Mini TTS 3 (also called Voxtral Speech 3; Artificial Analysis lists it as Voxtral 3 TTS)
- Size
- Voxtral Mini Transcribe 3 is 4B parameters (Mistral blog); other sizes not stated
- Access
- Mistral API (European, US and global endpoints); the blog also mentions self-deployment
- Weights
- Not stated; no Voxtral 3 repos on Hugging Face as of Oct 20, 2026unsure
- Price (Mistral)
- Mini Transcribe 3 $0.0035/min, Small Transcribe 3 $0.0040/min, Mini Transcribe Realtime 3 $0.008/min, Speech 3 $13 per 1M characters
- Price (Artificial Analysis)
- Lists Voxtral 3 TTS at $13 per 1M characters
- Voice agent SDK
- Built on LiveKit: Transcribe listens, a reasoning model calls tools, Speech replies
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: Voxtral Small Transcribe 3 is state-of-the-art in multilingual transcription, diarization and context-biasing.
Lab claim: Voxtral Mini TTS 3 achieves state-of-the-art expressiveness in 35+ languages and 50+ dialects.
Lab claim: Pro Voice replicates a voice from up to 30 minutes of audio across every supported language.
Sources
LABMistral AI announcement on Xx.com/MistralAI/status/2112574615318876178 · Oct 20, 2026LABVoxtral 3 (Mistral AI blog)mistral.ai/news/voxtral-3/ · Oct 20, 2026PRESSArtificial Analysis thread on X (pricing and its own evaluations)x.com/ArtificialAnlys/status/2112587905222709432 · Oct 20, 2026Tracking continues until an independent board lists it.