whatmodel.sh / stt / Whisper small
Whisper small
↗ share
+ contribute
Whisper small is an open-weights speech-to-text model from OpenAI, released September 2022. Its published size is 0.244B parameters. Licence: apache-2.0.
Catalogue identity notes
Multilingual small checkpoint (openai/whisper-small@973afd2). Hugging Face mirror created 2022-09-26 with the Whisper release (arXiv 2212.04356). Multilingual checkpoint (openai-whisper name 'small'), distinct from the English-only small.en. Params from the Whisper README size table.
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 Whisper small, out of 176 with a score, then more from OpenAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 152 |  ESPnet Conformer RNN-T (no LM) | 40±4 | local | ||
| 153 |  STT En Conformer-CTC Small | 39±3 | local | ||
| 154 | Whisper small | 39±2 | local | ||
| 155 |  Moonshine Base | 38±3 | local | ||
| 156 | Whisper base.en | 35±3 | 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
other tests16 not in a task · 98% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LibriSpeech test-clean WER +1 | 3.3% ✎ | ▁···▁▁▂·▄▄██ beats 7/56 | 20% |
| Omnilingual ASR paper · FLEURS-81 test mean CER (Whisper-covered languages) | 50.8% ✎ | ▅···▃···▃▃██ beats 0/10 | 11% |
| FLEURS English test WER | 6.5% ✎ | ▁··▁▁▁▁▁▃▃█▁ beats 13/56 | 7.6% |
| Omnilingual ASR paper · Common Voice 22 (76 languages) test mean CER | 111.7% ✎ | ▃▂▂····▂▆··█ beats 0/10 | 7.5% |
| FLEURS Mandarin test error rate | 23.1% ✎ | ▂▂··▃▄▂▂·▆██ beats 2/17 | 7.2% |
| PESTE · Persian FLEURS (fa_ir) corpus CER | 26.0% ✎ | ▄··▄·▄▄▄··██ beats 2/8 | 7.2% |
| Open ASR Leaderboard – Chilean Spanish · 3-dataset average WER | 16.4% ✎ | ▃▂·····▂▂·█▂ beats 0/8 | 7.0% |
| Danish ASR Leaderboard · mean WER over 5 Danish test sets (RyeAI) | 46.9% ✎ | ▂▂···▃▂▅█▃▂▂ beats 3/16 | 5.7% |
8 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Betthupferl (German dialect ASR benchmark) · Standard German audio → Standard German transcript WER | 20.3% ✎ | ▄·█·▄▄▄▄·█▄▄ beats 4/10 | 4.4% |
| Open Universal Arabic ASR Leaderboard · 6-test-set average WER | 55.1% ✎ | ▃·▂▂▅█▅▃▂▅▂▂ beats 0/22 | 4.3% |
| verli STT benchmark · SPGISpeech 1,200-clip sample WER | 4.1% ✎ | ▂·····▄▂█▂▂▄ beats 1/11 | 4.1% |
| verli STT benchmark · Earnings-22 WER (2,741 utt.) | 14.9% ✎ | ▃·▃▅··▃▃▅▅█▃ beats 3/13 | 3.8% |
| TurkASR-Bench (7 Turkish sources, 5,200 utterances) · corpus WER | 21.7% ✎ | ▃··▃▃··▅··█▅ beats 3/9 | 3.5% |
| AfriSpeech-200 test (108 African accents) · WER | 39.1% ✎ | ███····█████ beats 4/7 | 2.4% |
| Betthupferl (German dialect ASR benchmark) · dialect audio → Standard German transcript WER | 47.4% ✎ | ▃█····▃▅▃·▅▃ beats 4/10 | 1.9% |
| TED-LIUM test WER (short-form) | 12.3% ✎ | ▂····▂▂▂█▃▃▂ beats 10/13 | 1.1% |
$ where to run it
Other ways to run it
version
- Run it as published · needs ~1.1 GB · from github.com/openai (official)
uvx --from openai-whisper whisper audio.mp3 --model small
Official weights and documentation: huggingface.co/openai ↗
$ how the score was worked out · transcribe.cpp model catalog… 34%, Omnilingual ASR: Open-Source… 18%, verliapp/stt-benchmark · inde… 8% and 8 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
$ 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