OpenAI: Whisper large-v3 Turbo
whisper-large-v3-turbo- Price
- From $0.000011
- Modalities
- Audio
About Whisper large-v3 Turbo
Whisper large-v3 Turbo is a faster variant of Whisper large-v3. It is the same model except that the decoder is pruned from 32 layers to 4, leaving about 809 million parameters instead of 1.55 billion. The model card reports faster inference at the expense of a minor quality drop. It covers 99 languages and is released under the MIT license.
Where it works well
- Four decoder layers instead of 32 make it noticeably faster than large-v3 on recorded audio.
- About 809 million parameters means a smaller memory footprint for self-hosting.
- Keeps large-v3's 99-language coverage, so multilingual workloads do not need a different model.
- MIT-licensed weights allow broad commercial use and modification.
When to choose another model
- The card notes a minor quality degradation against large-v3, so accuracy-critical transcripts may justify the full model.
- It is a transcription checkpoint without diarization or a live-streaming interface.
- OpenAI's GPT-4o transcription models report lower error rates on their own measurements.
Getting started
Create API key
Create a key in Console, then use it with every model on the platform.
Send your first request
Copy the example for your language and run it against the endpoint.
Speech to textTokenLab endpointPOST/v1/audio/transcriptionscurl -X POST "https://api.tokenlab.sh/v1/audio/transcriptions" \ -H "Authorization: Bearer sk-xxx" \ -F file="@audio.mp3" \ -F operation="stt" \ -F model="whisper-large-v3-turbo" \ -F language="en" \ -F response_format="json"
Parameters
Speech to text
| Parameter | Values and limits | Default | Notes |
|---|---|---|---|
file | Inputs: ≥ 1, ≤ 1 | Not provided | Send one audio file in multipart/form-data. Maximum file size is 25 MiB. |
language | Not provided | Not provided | This field is accepted. The current public contract does not specify a complete value range or list. No additional values are promised. The omitted-field default is not verified. |
Pricing
The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.
Speech to text
per second- Official price
- $0.0000111
- Verified price
- —
- Discount
- —
| Official priceper second | Verified priceper second | Discount | |
|---|---|---|---|
| Speech to text | $0.0000111 | — | — |
Usage & activity
Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.
Usage & availability
Last 24 hours- Requests
- Success rate
- P95 latency
- Total tokens
Data is based on aggregate user requests, excluding status checks.
Open in Console
Open Whisper large-v3 Turbo in Console with a prompt ready to edit or send.
Help me try whisper-large-v3-turbo with a short audio request at /v1/audio/transcriptions. Show the result and cost.
Use cases
Fast batch transcription
Work through large audio backlogs where turnaround time matters more than the last point of accuracy.
Near-live pipelines
Process short chunks quickly enough for rolling captions in an app you host.
Cost-conscious self-hosting
Run the model on smaller GPUs than the full large-v3 requires, which helps when hardware budget is the limit.
First-pass drafts
Create a quick first transcript that human editors then correct, instead of transcribing from nothing.
Prompt examples
Transcribe this batch of two hundred short support calls as plain text.
Transcribe this Spanish webinar quickly; a rough draft is fine.
Transcribe this meeting recording into text I can proofread.
FAQ
What is Whisper large-v3 Turbo?
It is a pruned and fine-tuned version of Whisper large-v3 with a 4-layer decoder instead of 32. It is faster and smaller, with about 809 million parameters, at the cost of a minor quality drop.
Is it better than large-v3?
It is faster, not more accurate. Use it when throughput matters, and use whisper-large-v3 when you want the full decoder for maximum accuracy and can accept slower processing and a larger memory footprint.
How many languages does it support?
The model card lists 99 languages for multilingual speech recognition, the same coverage as the larger checkpoint, so you do not lose language support by choosing the faster variant.
Which license applies to Whisper large-v3 Turbo?
The model card on Hugging Face lists the MIT license, which permits commercial use. As with any open checkpoint, confirm the current license text on the card before you redistribute or build a product on it.
Does it label speakers?
No. It outputs a plain transcript with no speaker labels. For speaker attribution on recordings use gpt-4o-transcribe-diarize, or run a separate diarization step in your own pipeline.
How much does Whisper large-v3 Turbo cost?
On TokenLab, Whisper large-v3 Turbo costs $0.0000111 per second. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.
Which endpoint should Whisper large-v3 Turbo use?
Use https://api.tokenlab.sh/v1/audio/transcriptions for Whisper large-v3 Turbo. The request example below shows the matching code shape.
Which operations does Whisper large-v3 Turbo support?
Whisper large-v3 Turbo supports Speech to text. Select an operation above to see its endpoint and request example.
Compare Whisper large-v3 Turbo
Sources
Reviewed Oct 2, 2026
