Voxtral Small Transcribe 3

Use it to transcribe multilingual audio, identify speakers, and improve recognition of domain-specific terms.

Availability

Available today through Mistral’s EU-hosted API or for self-deployment

Pricing model

Not established

Deployment

EU-hosted API or self-deployment on own infrastructure

Privacy

Organizations control where audio is processed, how data is handled, and whether conversations are retained

Licence

Not established

• The model supports transcription across 50 languages. ^[1]

  The speech model implicitly interprets the semantic meaning of context better, producing natural speech across a diverse range of emotions and languages.

• It supports diarization and context biasing. ^[1]

  It achieves high-quality transcriptions at sub-500ms latency in over 35 languages, including with diarization and context biasing.

• It detects when a speaker has finished speaking. ^[1]

  It detects when someone has finished speaking so agents can respond without delay (end-of-turn detection).

• It is available through Mistral’s EU-hosted API or for self-deployment.

• A toolkit for voice agents that listen, respond, and take action Weekly brief 2026-W43