GPT-4o optimized for transcription with improved WER
98 languages · benchmarked
GPT-4o Transcribe is a speech-to-text model from OpenAI. In our standardized benchmarks it reaches 14.76% word error rate, ranking #14 of 35 models tested. It supports 98 languages and runs at $0.006/min — mid-range on price.
Its nearest benchmarked alternative is Soniox STT (Async). Best suited for multilingual workloads.
Overall Score
74.0
#14 of 35
Word Error Rate
14.76%
Character Error Rate
11.60%
Match Error Rate
14.32%
Word Info Lost
18.94%
Avg Latency
2.2s
Benchmarks Run
35
last 30 runs
Word error rate
17.22%
↓ 48.4% · 30d
3 runs ago
latest
Avg latency
2.5s
↓ 10.0% · 30d
3 runs ago
latest
overall WER · 35 models
GPT-4o Transcribe
field · 34 models
lower-left is better
8 categories
WER
lower is better
Legal
3.6%
Finance
6.4%
General
8.3%
Medical
11.6%
Technical
18.2%
Conversational
25.8%
Noisy Environment
34.6%
Code-Switching
44.8%
| Category | WER | CER | MER | WIL | Latency | Benchmarks |
|---|---|---|---|---|---|---|
Code-Switching | 44.79% | 41.90% | 44.78% | 47.00% | 2.6s | 2 |
Conversational | 25.80% | 20.98% | 25.80% | 29.79% | 2.2s | 2 |
Finance | 6.37% | 3.32% | 6.30% | 10.21% | 3.9s | 2 |
General | 8.29% | 6.78% | 8.19% | 12.74% | 2.1s | 19 |
Legal | 3.65% | 2.81% | 3.65% | 4.34% | 2.5s | 2 |
Medical | 11.62% | 9.48% | 11.62% | 13.04% | 2.1s | 2 |
Noisy Environment | 34.58% | 22.58% | 31.25% | 43.22% | 1.5s | 4 |
Technical | 18.22% | 15.04% | 18.15% | 19.56% | 2.7s | 2 |
5 accents
| Accent | WER | Benchmarks |
|---|---|---|
African | 26.15% | 1 |
Australian | 12.31% | 1 |
Indian | 26.98% | 1 |
British | 4.48% | 1 |
American | 18.87% | 1 |
Rate
$0.006
/min
Per-second billing. Bring your own provider key and pay your provider directly — a 5% routing fee applies (first 100 min/mo free).
Set up BYOKCost estimator
1 hour
$0.36
10 hours
$3.60
100 hours
$36
1,000 hours
$360
Billed per second of audio processed.
Model ID
openai/gpt-4o-transcribe
Authenticate every request with your secret API key as a Bearer token. Issue a key from your dashboard.
Use it with your agent
Paste this into Claude Code, Cursor, Codex or any agent with a shell. It installs the CLI and the OpenTranscription skill, then transcribes with this model.
Quickstart
Upload your audio, create a job with this model, then poll the job or set a webhook_url to be notified when it completes.
Key parameters
file_path
string
required
Storage path returned by the upload step.
model
string
required
The model to run this job on.
models
string[]
Ordered fallback chain: primary first, then backups tried in order if a provider fails.
language
string
ISO 639-1 language code. Omit to auto-detect.
word_timestamps
boolean
Per-word start, end and confidence. Defaults to true; send false to leave them out.
webhook_url
string
HTTPS URL notified with the result when the job completes. Delivered events are signed (X-OT-Signature).
title
string
Display name for this job in the dashboard. Falls back to the file name.
Response
A completed job returns the transcript in the OpenTranscription Unified Schema (OTUS).
Webhooks — skip polling
We POST a signed event to your webhook_url when a job completes or fails; verify the X-OT-Signature (HMAC-SHA256, reject if older than 300 s) and dedupe on the event id, then fetch the full transcript via GET /api/v1/transcriptions/{id}.
Rate-limited per tier — see the X-RateLimit-* response headers.
Full API referenceSupported Languages
Supported Formats
Features
Limits
Max file size
25 MB
Max duration
No documented limit