Mistral Releases Voxtral Transcribe 2: State-of-the-Art Speech-to-Text with Open Weights
Published 2026-02-04Foundation ModelsMedium
Summary
Mistral released Voxtral Transcribe 2, a family of two speech-to-text models: Voxtral Mini Transcribe V2 for batch transcription and Voxtral Realtime for live applications requiring low latency. Voxtral Realtime is released under Apache 2.0 open-weights license and is deployable on edge devices for privacy-first applications. The release includes an updated audio playground in Mistral Studio for instant testing with diarization and timestamps. Voxtral Mini Transcribe V2 achieves approximately 4
Alignment: New signal not yet covered
mistralvoxtralspeech-to-texttranscriptionvoice-agentsopen-weightsapache-2diarizationreal-timemultilingualedge-aienterprise