whatmodel.sh / stt / wav2vec 2.0 Large 960h
wav2vec 2.0 Large 960h
↗ share
+ contribute
wav2vec 2.0 Large 960h is an open-weights speech-to-text model from Meta, released June 2020. Its published size is 0.32B parameters. Licence: apache-2.0.
Catalogue identity notes
Exact checkpoint facebook/wav2vec2-large-960h (CTC fine-tuned on LibriSpeech 960h). Params: fp32 pytorch_model.bin 1,262,009,187 bytes / 4; released: arXiv 2006.11477v1 month.
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 wav2vec 2.0 Large 960h, out of 176 with a score, then more from Meta.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 170 | wav2vec 2.0 Conformer RoPE Large 960h | 13±4 | local | ||
| 171 | data2vec Audio Large 960h | 13±4 | local | ||
| 172 | wav2vec 2.0 Large 960h | 7±3 | local | ||
| 173 | data2vec Audio Base 960h | 4±4 | local | ||
| 174 | wav2vec 2.0 Base 960h | 2±3 | local | ||
| more from Meta | |||||
| 1 | Muse Voice Transcribe 1.0 | 71±3 | api $0.0030/audio min | ||
| 58 | Omnilingual ASR LLM 7B | 58±2 | local | ||
| 59 | Omnilingual ASR LLM 7B v2 | 58±3 | local | ||
see all speech-to-text models · find the ones that fit your machine
$ what it's good at
other tests6 not in a task · 96% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Open ASR English short-form WER (March 2026) | 26.8% ✎ | ▁▁▃▃··▁▁▂██▆ beats 3/86 | 37% |
| LibriSpeech test-other WER lab | 6.3% ✎ | ▂·▂·▃······█ beats 2/6 | 33% |
| LibriSpeech test-clean WER +1 | 3.0% ✎ | ▂··▂·▃▂█▂▃▃▃ beats 11/16 | 14% |
| AfriSpeech-200 test (108 African accents) · WER | 71.7% ✎ | ███····█████ beats 0/7 | 5.9% |
| Speech Accessibility Project dysarthric speech · zero-shot WER (rated subset, N=11,218) | 39.3% ✎ | ▃·▂··▂····▆█ beats 0/9 | 4.3% |
| TED-LIUM test WER (short-form) | 16.5% ✎ | ▂····▂▃▂█▃▃▂ beats 2/13 | 2.7% |
$ where to run it
Other ways to run it
- Run it as published · needs ~2.04 GB · from huggingface.co/facebook
uv run --with transformers --with torch python -c "from transformers import pipeline; print(pipeline('automatic-speech-recognition', model='facebook/wav2vec2-large-960h', chunk_length_s=30)('audio.wav')['text'])"
Weights and documentation: huggingface.co/facebook ↗
$ how the score was worked out · Meta (facebook org) Hugging F… 44%, Open ASR Leaderboard 37%, AfriSpeech-200 (arXiv:2310.00… 6% and 2 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 |
|---|---|---|
 Meta (facebook org) Hugging Face ASR model cards: published LibriSpeech WER for its own checkpoints · lab-reported | 2 | 44% |
| Open ASR Leaderboard | 1 | 37% |
 AfriSpeech-200 (arXiv:2310.00274v1) · zero-shot WER of open ASR models on African-accented English | 1 | 6% |
| Speech Robust Bench (arXiv:2403.07937v3) · clean-speech WER of 25+ open ASR models | 2 | 5% |
| Dysarthric ASR with clinical context on the Speech Accessibility Project (arXiv:2605.02782v2) | 1 | 4% |
| prior · pull toward the middle | 4% | |
| weighted average | = 7.0 |
$ 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