whatmodel.sh / stt / ESPnet wav2vec 2.0 Large + Conformer + Transformer LM

ESPnet wav2vec 2.0 Large + Conformer + Transformer LM

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

ESPnet wav2vec 2.0 Large + Conformer + Transformer LM is an open-weights speech-to-text model from ESPnet, released December 2021. Licence: cc-by-4.0.

Catalogue identity notes

Exact LibriSpeech 960h recipe run, README lines 171–208. Published checkpoint and inference API verified in the model card. Decoder and checkpoint averaging are recorded per result. Architecture names do not denote a generic pretrained model. Local command is the publisher’s CPU inference example for audio.wav; to reproduce the reported WER use the exact recipe decoder settings recorded with each result. Memory is an estimated 8GB runtime budget for short utterances, not a measurement.

45.1±3.9#122 of 162 speech-to-text modelsfrom 2 benchmark results across 1 source* thinly tested, treat the rank as provisional

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 ESPnet wav2vec 2.0 Large + Conformer + Transformer LM, out of 176 with a score, then more from ESPnet.

#modelscorewhereRAMreleased
133SenseVoice Small Alibaba (FunAudioLLM)46±3local~0.3GB2024-07
134STT En FastConformer-CTC Large NVIDIA45±3local~1.1GB2023-06
135ESPnet wav2vec 2.0 Large + Conformer + Transformer LM ESPnet45±4local~8GB2021-12
136ESPnet Conformer + S4 decoder (no LM) ESPnet45±4local~8GB2022-12
137Omnilingual ASR CTC 300M Meta45±2local~1.3GB2025-11
more from ESPnet
112ESPnet HuBERT Large + Conformer + Transformer LM ESPnet50±4local~8GB2022-01
113ESPnet WavLM Large + Conformer + Transformer LM ESPnet50±4local~8GB2022-01
129ESPnet E-Branchformer (no LM) ESPnet46±4local~0.8GB2023-01

see all speech-to-text models · find the ones that fit your machine

$ what it's good at

other tests2 not in a task · 94% of its score
testresultvs the others in its rundifficultycountswhy
LibriSpeech test-other WER lab5.1% ✎█▃▃·▃▅·▅···▅ beats 6/11avg 3.7%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 73%typical
LibriSpeech test-clean WER lab2.1% ✎█··▃▃▅▅▃···▅ beats 9/11avg 1.9%▇▇▇▇▇▇▇▇ 21%blunt

$ where to run it

Other ways to run it

  1. Run it as published · needs ~8 GB · from huggingface.co/espnet (official)
    uv run --python 3.10 --with espnet --with espnet_model_zoo --with librosa --with s3prl python -c "import librosa; from espnet2.bin.asr_inference import Speech2Text; asr=Speech2Text.from_pretrained(model_tag='espnet/simpleoier_librispeech_asr_train_asr_conformer7_wav2vec2_960hr_large_raw_en_bpe5000_sp', lm_weight=1); speech,_=librosa.load('audio.wav', sr=16000, mono=True); print(asr(speech)[0][0])"

Official weights and documentation: huggingface.co/espnet ↗

$ how the score was worked out · ESPnet LibriSpeech ASR recipe… 94%

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

sourceresultsshare of the scorelicence
ESPnet LibriSpeech ASR recipe results (pinned September 2026) · lab-reported2▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 94%Apache-2.0
prior · pull toward the middle▇▇ 6%
weighted average= 45.1

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

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