voxtral-mini-transcribe-realtime-3

Transcribed audio can be routed into classification workflows, but no classification benchmark or general-purpose reasoning capability is documented.

Realtime transcription can supply the transcript required for meeting summaries, although the profile provides no evidence of native summarisation or reasoning capabilities.

A 65,536-token context window supports processing substantial transcripts for downstream field extraction, but no structured-output or extraction capability is explicitly documented.

The large context window is suitable for extracting entities from long transcripts, though the model profile documents transcription rather than dedicated language-understanding features.

The profile identifies a realtime speech-transcription model and flags no vision capability, so it lacks evidence for document image OCR.

The model is identified for realtime transcription, with no documented voice-generation or speaker-cloning capability.

Transcription converts speech to text and does not provide evidence of text-to-speech or audio-generation capability.