Qwen Audio 3.1 ASR

POST /v1/audio/transcriptions

Speech-to-text for short clips and long recordings, with sentence segments, word timestamps, and multilingual and Chinese dialect recognition.

At a glance

Field	Value
Model id	`qwen-audio-3-1-asr`
Model release date	2026-09-10
Input modalities	Audio
Output modalities	Text
Context window	-
Weight precision	-
Region	Singapore
Features	transcription, speech_to_text, multilingual
Native inference	No
New	Yes
Supported endpoints	`POST /v1/audio/transcriptions`
Alternate model ids	`qwen-audio-3.1-asr`, `alibaba/qwen-audio-3-1-asr`

Pricing

Charge	Spec	Rate
Input	per 1M audio input tokens	$0.30
Output	per 1M generated tokens	$0.94

Example request

curl https://api.empiriolabs.ai/v1/audio/transcriptions \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -F model=qwen-audio-3-1-asr \
  -F file=@meeting.mp3

Parameters

This model accepts the standard chat completion parameters (see the API reference).

Notes

Sending audio

Send a file as multipart file, or JSON with audio_url or audio_base64, in common audio formats. Short clips and long recordings are both accepted.

Getting the result

The request returns a job_id and a poll_url right away. Poll GET /v1/jobs/<job-id> until status is completed. Short clips finish in seconds; long recordings take longer.

What comes back

The result of the completed job holds the full transcript as text, sentence segments with start and end times, and word timestamps. The language is detected automatically, including Chinese dialects.

Billing

Input and output tokens are priced per 1M, and each request is billed from the usage the model reports.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/qwen-audio-3-1-asr.