Voxtral Mini Transcribe 2
Mistral AI
Voxtral Mini Transcribe 2 (API id voxtral-mini-2602) is Mistral's batch speech-to-text model, described by the vendor as an efficient audio input model pre-trained and optimized for transcription purposes only. It is served from /v1/audio/transcriptions with speaker diarization, word-level timestamps and context biasing of up to 100 custom terms, accepts recordings up to 3 hours per request, and covers 13 languages. Mistral ships it in the Premier tier, so weights are not published, and prices it at $0.003 per minute.
Strengths
- Speaker diarization available directly in the batch transcription request
- Word-level timestamps returned from /v1/audio/transcriptions
- Context biasing with up to 100 custom terms for domain vocabulary
- Accepts recordings up to 3 hours per request, suitable for meetings and call archives
- Vendor claims noise-robust accuracy across 13 languages
- Listed at $0.003 per minute, among the cheaper hosted transcription rates
Caveats
- Batch only: live transcription requires the separate Voxtral Realtime model (voxtral-mini-transcribe-realtime-2602), and Mistral notes realtime transcription is not compatible with the diarize parameter.
- 13 languages is a narrow footprint next to Whisper-class models, and Mistral does not enumerate which 13 on the audio overview page.
- Premier tier means closed weights, unlike its Apache-2.0 siblings Voxtral Mini Transcribe Realtime and Voxtral Small - no self-hosting or VPC deployment of this checkpoint.
- Transcription only: the model page lists just the Transcriptions and Timestamps features, with no audio understanding, chat or function calling.
- Mistral publishes no benchmark numbers as text for this model; the model page shows Speed, Performance and Context as '--'.
- Vendor
- Mistral AI
- License
- Proprietary
- Release Date
- 2026-02-04
- Modalities
- audiotext
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
$0.003 per minute of audio via the Mistral API, as listed on the model page. Premier tier, so there is no self-hosted or open-weights option.