whatmodel.sh / stt / GPT-4o Mini Transcribe
GPT-4o Mini Transcribe
↗ share
+ contribute
GPT-4o Mini Transcribe is a closed-weights speech-to-text model from OpenAI, released March 2025. Licence: proprietary.
Score within the Open ASR English short-form WER (March 2026) comparison set only.
$ how it compares
The speech-to-text models ranked either side of GPT-4o Mini Transcribe, out of 176 with a score, then more from OpenAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 72 | Zipformer Transducer XL 290M | 56±5 | comparison only | ||
| 73 |  Parakeet TDT-CTC 1.1B | 56±4 | local | ||
| 74 | GPT-4o Mini Transcribe | 56±2 | api $0.0030/audio min | ||
| 75 |  Omnilingual ASR LLM 1B v2 | 56±3 | local | ||
| 76 | Omnilingual ASR LLM 1B | 56±2 | local | ||
| more from OpenAI | |||||
| 13 | GPT Transcribe | 64±4 | api $0.0045/audio min | ||
| 48 | GPT-Realtime-Whisper | 59±3 | api $0.017/audio min | ||
| 61 | Whisper large-v3-turbo | 57±1 | api $0.0005/audio min +local | ||
see all speech-to-text models · find the ones that fit your machine
$ what it's good at
long-form transcription█·····█···██ #3 of 4 * · 7.0% of its score · 1 test
Beats Kyutai STT 2.6B (en) here, though they rank higher overall. Trails Voxtral Mini 4B Realtime (2602) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Voxtral report · English long-form 4-set WER | 8.0% ✎ | ▄··█·······▄ beats 0/3 | 7.0% |
multilingual transcription▂··▃▄█▆▅▄▂·▂ #31 of 36 · 18% of its score · 3 tests
Trails Omnilingual ASR CTC 1B v2, Omnilingual ASR CTC 7B v2, Voxtral Mini 4B Realtime (2602) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Voxtral report · Common Voice 12-language error rate | 12.8% ✎ | █··█·······█ beats 1/2 | 7.1% |
| Voxtral report · FLEURS 13-language error rate | 8.0% ✎ | █·█········█ beats 1/2 | 6.5% |
| mu-bench voice-agent caller utterances · batch UER (5 locales) | 27.2% ✎ | █·····▄····▄ beats 1/3 | 4.7% |
other tests12 not in a task · 73% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SPGISpeech test WER | 4.5% ✎ | ▄▄··▄██··▄██ beats 0/11 | 10% |
| SwitchBoard WER | 17.3% ✎ | ▄······▄···█ beats 0/3 | 8.7% |
| CHiME-4 WER | 11.3% ✎ | █··█·█·····█ beats 0/3 | 8.1% |
| Earnings-21 10-minute long-form WER | 10.1% ✎ | █···█·····██ beats 0/3 | 7.5% |
| Voxtral report · English short-form 10-set WER | 7.9% ✎ | █··█·█·····█ beats 2/3 | 7.4% |
| Earnings-22 10-minute long-form WER | 12.3% ✎ | █····█·█···█ beats 2/3 | 7.3% |
| GigaSpeech test WER | 14.8% ✎ | ▁····▃··▁▄█▂ beats 1/17 | 7.1% |
| LibriSpeech test-other WER | 4.7% ✎ | ▁▁▁▂▅▃▅▄█▂▂▁ beats 27/54 | 6.2% |
4 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Danish ASR Leaderboard · mean WER over 5 Danish test sets (RyeAI) | 22.2% ✎ | ▂▂···▃▂▅█▃▂▂ beats 13/16 | 4.4% |
| VoxPopuli English test WER | 7.3% ✎ | █▄▄··████▄██ beats 4/16 | 3.7% |
| LibriSpeech test-clean WER | 1.9% ✎ | ▁·▁·▁▁▁▁▃▇█▆ beats 60/100 | 1.9% |
| FLEURS English test WER | 3.2% ✎ | ▁··▁▁▁▁▂▃█▃▁ beats 63/64 | 1.6% |
$ where to run it
Through an API
api · $0.0030/audio min · developers.openai.com ↗ · price sourceProvider-published estimated cost per audio minute; underlying billing uses audio/text tokens. Exact endpoint gpt-4o-mini-transcribe. Product: Audio transcription. API model: gpt-4o-mini-transcribe. Settings: endpoint=/v1/audio/transcriptions, streaming=no. Price basis: provider-published estimate. Currency: USD.
$ how the score was worked out · Mistral AI — Voxtral technica… 62%, Voxtral Realtime 28%, mu-bench: A Multilingual Utte… 5% and 1 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Mistral AI — Voxtral technical report | 10 | 62% |
| Voxtral Realtime | 4 | 28% |
| mu-bench: A Multilingual Utterance Transcription Benchmark (arXiv:2609.32082v1), Table 5 | 1 | 5% |
 Danish Speech-to-Text Leaderboard (RyeAI) · results dataset | 1 | 4% |
| prior · pull toward the middle | 1% | |
| weighted average | = 55.6 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json