Voxtral Mini Transcribe 2
Voxtral Mini Transcribe 2 is the batch model in the February 2026 Transcribe 2 release. It adds speaker diarization, word-level timestamps and context biasing for recorded audio.
Deprecated API; check access details
Recorded audio and vocabulary hints
The launch supports recordings up to three hours and up to 100 bias words or phrases. Biasing is optimized for English; other languages are experimental. Speaker labels have a limit: overlapping speech typically produces one speaker’s transcript.Voxtral Mini Transcribe 2 release announcement
Batch and realtime are different releases
This version uses voxtral-mini-2602. The simultaneously released Realtime model has a streaming architecture and Apache weights. The batch model is a proprietary service with no public checkpoint listed; the Realtime license does not transfer to it.Voxtral Mini Transcribe 2 model and lifecycle recordVoxtral Mini Transcribe 2 release announcement
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Deprecated API; check access details |
| Launch date | |
| Access | API |
| Model identifiers | voxtral-mini-2602Voxtral Mini Transcribe V2 |
| Weight access | No public weight download is listed in the official model record.Voxtral Mini Transcribe 2 model and lifecycle record |
| Access and lifecycle | Hosted API deprecated on 31 October 2026.Voxtral Mini Transcribe 2 model and lifecycle record |