Inference APIs
Models/Transcription/Voxtral Small 24B

Voxtral Small 24B API

Available New

Voxtral Small (July 2025 release) is Mistral's speech-understanding model built on Mistral Small 3. Here it is served as speech to text on the OpenAI-compatible transcriptions endpoint, with automatic language detection across English, Spanish, French, Portuguese, Hindi, German, Dutch and Italian. Apache 2.0 weights.

Try in playground
Speech to text8 languagesAuto languageText only
Modality
Audio → Text
Price
$0.004 / audio minute
Speed
0.09× real time
Model ID
mistralai/Voxtral-Small-24B-2507

Quickstart

You need an API key to call this model. Create a free account or log in.

Set INFERENCE_API_KEY to your key from the API Keys page, then run:

curl https://api.inferenceapis.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $INFERENCE_API_KEY" \
  -F "model=mistralai/Voxtral-Small-24B-2507" \
  -F "file=@meeting.mp3" \
  -F "response_format=verbose_json"
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.inferenceapis.com/v1",
    api_key=os.environ["INFERENCE_API_KEY"],
)
with open("meeting.mp3", "rb") as f:
    result = client.audio.transcriptions.create(
        model="mistralai/Voxtral-Small-24B-2507",
        file=f,
        response_format="verbose_json",
    )
print(result.text)
import fs from "node:fs";
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.inferenceapis.com/v1",
  apiKey: process.env.INFERENCE_API_KEY,
});
const result = await client.audio.transcriptions.create({
  model: "mistralai/Voxtral-Small-24B-2507",
  file: fs.createReadStream("meeting.mp3"),
  response_format: "verbose_json",
});
console.log(result.text);

When to use it

A good fit when
  • Accuracy on messy real-world audio in English, Spanish, French, Portuguese, Hindi, German, Dutch or Italian
  • You want Mistral's Voxtral without running a 24B model yourself
Look elsewhere when
  • You need timestamps or subtitles (use Whisper or Parakeet)
  • Languages outside its eight; use Whisper
  • Speaker diarization

What it costs in practice

WorkloadCost
1 hour of audio$0.24
100 hours of audio$24.00
1,000 hours of audio$240

Computed from the live rate below. There is no subscription or minimum; new accounts start with free credit.

Endpoint

POST https://api.inferenceapis.com/v1/audio/transcriptions
Headers Authorization: Bearer $INFERENCE_API_KEY
Content-Type: multipart/form-data
Compatible OpenAI wire format. /openai/v1/… is accepted too, so Groq-style base URLs work with only the host changed.

Model IDs and aliases

Send any of these as model; they all resolve to this model and bill at its rate.

IDNote
mistralai/Voxtral-Small-24B-2507Canonical
voxtralAlias
voxtral-smallAlias
voxtral-small-24bAlias

Request body

ParameterTypeDescription
model required string Model ID: mistralai/Voxtral-Small-24B-2507 — also accepted: voxtral, voxtral-small, voxtral-small-24b
file required file The audio file (multipart/form-data): mp3, mp4, m4a, wav, webm, flac, ogg. Up to 100 MB.
language string ISO-639-1 language code, e.g. en. Improves accuracy and latency when known.
prompt string Optional text to guide style or spelling of names and terms.
response_format string json (default), verbose_json, text, srt or vtt.
timestamp_granularities array ["segment"] and/or ["word"] with verbose_json.
temperature number Sampling temperature between 0 and 1.

Response

200 OK · application/json
{
  "language": "en",
  "duration": 12.4,
  "text": "Thanks everyone for joining. Let's start with the roadmap.",
  "segments": [
    { "id": 0, "start": 0.0, "end": 3.1, "text": "Thanks everyone for joining." },
    { "id": 1, "start": 3.1, "end": 6.0, "text": "Let's start with the roadmap." }
  ]
}

Pricing

Price$0.004 / audio minute
BillingPay as you go from prepaid credits. No subscription, no daily request or token cap. See all pricing.

Notes

  • Billing is per minute of audio, measured from the uploaded file.
  • Text only: verbose_json adds the duration but no timestamps or language field, and srt/vtt return a 400 pointing you to Whisper or Parakeet.

Compare Voxtral Small 24B with…

Whisper Large v3Parakeet TDT 0.6B v3Voxtral Mini 3BOpenAI Whisper (whisper-1)Whisper Large v3 Turbo on GroqDeepgram Nova-3AssemblyAI Universal

Languages

Per-language codes, tips and quickstarts: speech-to-text languages.

Questions about Voxtral Small 24B

Which languages does it support?

Mistral lists English, Spanish, French, Portuguese, Hindi, German, Dutch and Italian, with automatic language detection. Send language to pin one. For other languages use Whisper, which covers about a hundred.

Voxtral or Whisper?

Voxtral is a 24B language model that listens; Whisper is a dedicated speech model. Mistral's own benchmarks put Voxtral ahead of Whisper large-v3 on its eight languages; Whisper covers far more languages and, here, returns timestamps, which Voxtral does not. Voxtral Small costs twice Whisper per minute here.

Does it return timestamps?

No. On this host Voxtral returns the transcript and the duration only, with no segment or word timestamps and no language field, which we confirmed on a 23-second clip. Asking for srt, vtt or timestamp_granularities returns a 400 that names the models that do: Whisper and Parakeet.

Can it do speaker diarization or answer questions about the audio?

Not on this endpoint. It is served as transcription only.

Licence

Voxtral Small 24B is published by Mistral AI under the Apache 2.0. Your use of its output is also subject to that licence.

Errors

Errors use the OpenAI envelope: {"error": {"message", "type", "code"}}. 401 missing or invalid key · 402 insufficient_balance · 404 model_not_found · 503 model_unavailable · 502 backend error, safe to retry. Full table with what to retry: API documentation. Errors from other providers: provider error reference.