# Voxtral Mini 4B Realtime

Mistral's streaming speech recognition: live audio over /v1/realtime, a transcript while the speaker talks, 480 ms behind.

Model id: `voxtral-mini-4b-realtime`
Quantization: bf16
Regions: ATL
Hugging Face: [mistralai/Voxtral-Mini-4B-Realtime-2602](https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602)

## Pricing

- Input: $0.000133 / second

## Limits

Per subscription, on this model:

- Transcription: 30 requests a minute, 5 at once

## Inputs

- audio: sources base64; formats wav, mpeg; max length 14K seconds; max size 25 MB

## Output: transcription

| Parameter | Type | Values | Default |
| --- | --- | --- | --- |
| `language` | any |  |  |
| `prompt` | any |  |  |
| `temperature` | number | 0–1 |  |
| `response_format` | string | json, text |  |

## Endpoints

- [`POST /v1/audio/transcriptions`](/docs/reference/create-transcription.md): Create Transcription