Voxtral Mini 2507
Voxtral Mini 2507 adds audio input to Ministral 3B. Its Apache 2.0 checkpoint can transcribe speech, translate it and answer questions about a recording.
Deprecated API; check access details
Audio understanding in a compact checkpoint
The card describes a 32K context with up to 30 minutes of audio for transcription or 40 minutes for understanding. It retains the text abilities of Ministral 3B and can request function calls from spoken instructions.Voxtral Mini 2507 official model card
Audio conversation format
The release supports multiple audio clips in a message and audio across multiple user turns. Its card notes that system prompts are not supported. The downloadable understanding checkpoint should be distinguished from the separate transcription-optimized model used by the hosted transcription route.Voxtral Mini 2507 official model cardVoxtral Mini 2507 model and lifecycle record
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Deprecated API; check access details |
| Launch date | |
| Access | API, Downloadable |
| Model identifiers | Voxtral-Mini-3B-2507voxtral-mini-2507 |
| Context window | 32,000 tokens The model catalog advertises 32k context for this version. This is not a measured recall limit. Voxtral Mini 2507 model and lifecycle record |
| Weight license | Apache 2.0Voxtral Mini 2507 official model card |
| Access and lifecycle | Hosted API deprecated on 27 February 2026. Downloadable weights remain listed.Voxtral Mini 2507 model and lifecycle recordVoxtral Mini 2507 official model card |
| Language-model base | Ministral 3BVoxtral Mini 2507 official model card |