whatmodel.sh / stt / Whisper large-v2
Whisper large-v2
↗ share
+ contribute
Whisper large-v2 is an open-weights speech-to-text model from OpenAI, released December 2022. Its published size is 1.55B parameters. Licence: apache-2.0.
Catalogue identity notes
Exact checkpoint openai/whisper-large-v2 (openai-whisper name 'large-v2'). Params from the Whisper README size table. Licence as declared on the HF card.
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 large-v2, out of 176 with a score, then more from OpenAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 103 |  OWSM-CTC v4 1B | 52±4 | local | ||
| 104 |  Omnilingual ASR CTC 1B | 51±2 | local | ||
| 105 | Whisper large-v2 | 51±2 | local | ||
| 106 |  STT En Conformer-CTC Large | 51±4 | local | ||
| 107 | Omnilingual ASR LLM 300M v2 | 51±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
long-form transcription▁▁▄▁·▁▂█▄▁▁▁ #12 of 27 * · 17% of its score · 1 test
Beats Chirp 3, Parakeet TDT 1.1B, Parakeet RNNT 1.1B here, though they rank higher overall. Trails Whisper large (v1) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Open ASR English long-form WER | 9.4% ✎ | ▁▁▅▁·▃▅█▆▁▂▂ beats 16/30 | 17% |
other tests10 not in a task · 81% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LibriSpeech test-clean WER +1 | 2.7% ✎ | ▁···▁▁▂·▄▄██ beats 14/56 | 22% |
| Open ASR English short-form WER (March 2026) | 7.8% ✎ | ▁▁▁▃··▁▁▂██▆ beats 40/86 | 17% |
| FLEURS English test WER | 4.2% ✎ | ▁··▁▁▁▁▁▁▃█▁ beats 47/56 | 8.6% |
| FLEURS Mandarin test error rate | 15.4% ✎ | ▂▂··▂▄▃▂·▆██ beats 5/17 | 8.1% |
| PESTE · Persian FLEURS (fa_ir) corpus CER | 8.5% ✎ | ▄··▄·▄▄▄··██ beats 6/8 | 8.1% |
| Betthupferl (German dialect ASR benchmark) · Standard German audio → Standard German transcript WER | 12.2% ✎ | ▄·█·▄▄▄▄·█▄▄ beats 7/10 | 4.9% |
| Open Universal Arabic ASR Leaderboard · 6-test-set average WER | 40.2% ✎ | ▃·▂▂▅█▅▃▂▅▂▂ beats 7/22 | 4.9% |
| TurkASR-Bench (7 Turkish sources, 5,200 utterances) · corpus WER | 15.3% ✎ | ▃··▃▃··▅··█▅ beats 7/9 | 4.0% |
2 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Betthupferl (German dialect ASR benchmark) · dialect audio → Standard German transcript WER | 39.5% ✎ | ▃█····▃▅▃·▅▃ beats 7/10 | 2.2% |
| TED-LIUM test WER (short-form) | 12.1% ✎ | ▂····▂▂▂█▃▃▂ beats 11/13 | 1.2% |
$ where to run it
Other ways to run it
version
- Run it as published · needs ~4.22 GB · from github.com/openai (official)
uvx --from openai-whisper whisper audio.mp3 --model large-v2
Official weights and documentation: huggingface.co/openai ↗
$ how the score was worked out · transcribe.cpp model catalog… 38%, Open ASR Leaderboard 34%, PESTE · Persian Speech to Tex… 8% and 4 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 |
|---|---|---|
 transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports | 3 | 38% |
 Open ASR Leaderboard | 2 | 34% |
| PESTE · Persian Speech to Text benchmark (fleurs-fa-ir-v1) | 1 | 8% |
| German dialect ASR Benchmark (Betthupferl; kamaal-08/bavarian-asr-leaderboard) | 2 | 7% |
| Open Universal Arabic ASR Leaderboard (Elm Research Center) | 1 | 5% |
| Turkish ASR Leaderboard (TurkASR-Bench, yagmurtuncer) | 1 | 4% |
 Speech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models | 2 | 2% |
| prior · pull toward the middle | 2% | |
| weighted average | = 51.1 |
$ 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