Voxtral TTS 2603
Voxtral TTS 2603 is the March 2026 speech-generation release. Its BF16 weights and reference voices are available under CC BY-NC 4.0. Mistral's serverless model is deprecated in favor of TTS 3. The downloadable license does not establish commercial-use rights.
Deprecated API; check access details
Understand why the voices affect the license
Mistral states that the model inherits its noncommercial license from the supplied reference voices. The card lists nine languages, 20 preset voices, streaming and batch inference, and 24kHz output. Supported formats include WAV, PCM, FLAC, MP3, AAC and Opus.Voxtral TTS 2603 official weight card
Separate the deprecated API from weights
The model page records API deprecation on 31 October 2026. The official weight repository remains available. API access is subject to Mistral's retirement schedule; a published model identifier does not guarantee continued service after retirement.Voxtral TTS 2603 API lifecycle pageVoxtral TTS 2603 official weight card
Evaluate the complete task
Practical guidance. For a local research prototype, compare voice consistency and pronunciation in each required language. Before moving that prototype into a commercial product, choose an access route with terms that permit the intended use. API service terms and downloadable weight terms require separate checks.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Deprecated API; check access details |
| Launch date | |
| Access | API, Downloadable |
| Model identifiers | voxtral-mini-tts-2603Voxtral-4B-TTS-2603 |
| API status | Deprecated on 31 October 2026Voxtral TTS 2603 API lifecycle page |
| Weight license | CC BY-NC 4.0; noncommercial weightsVoxtral TTS 2603 official weight card |
| Weight repository | mistralai/Voxtral-4B-TTS-2603Voxtral TTS 2603 official weight card |
Related models
- Voxtral TTS 3
Mistral AI ยท Voxtral
A preview speech-generation model with voice cloning, emotion tags and controls for speaking pace.
Launch date PreviewAPI