whatmodel.sh / stt / Omnilingual ASR LLM 7B
Omnilingual ASR LLM 7B
↗ share
+ contribute
Omnilingual ASR LLM 7B is an open-weights speech-to-text model from Meta, released November 2025. Its published size is 7.8B parameters. Licence: apache-2.0.
Catalogue identity notes
Exact checkpoint facebook/omniASR-LLM-7B (fairseq2 card 'omniASR_LLM_7B', first Omnilingual ASR release, Nov 2025); distinct from its '_v2' successor. Params: card table. Needs libsndfile (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 Omnilingual ASR LLM 7B, out of 176 with a score, then more from Meta.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 56 |  Canary-1B-v2 | 58±1 | local | ||
| 57 |  Phi-4-multimodal-instruct | 58±2 | local | ||
| 58 | Omnilingual ASR LLM 7B | 58±2 | local | ||
| 59 | Omnilingual ASR LLM 7B v2 | 58±3 | local | ||
| 60 |  Distil-Whisper distil-large-v3.5 | 58±1 | local | ||
| more from Meta | |||||
| 1 | Muse Voice Transcribe 1.0 | 71±3 | api $0.0030/audio min | ||
| 62 | Omnilingual ASR LLM 3B | 57±2 | local | ||
| 75 | Omnilingual ASR LLM 1B v2 | 56±3 | local | ||
see all speech-to-text models · find the ones that fit your machine
$ what it's good at
multilingual transcription▂··▃▄█▆▅▄▂·▂ #13 of 36 * · 60% of its score · 1 test
Beats Cohere Transcribe Arabic (07-2026), Qwen3-ASR-1.7B, Voxtral Mini Transcribe V2 here, though they rank higher overall. Trails SeamlessM4T v2 Large, Omnilingual ASR LLM 1B v2, Whisper large-v3 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Open ASR multilingual WER | 4.7% ✎ | ▂··▂▃▂█▃█▂▁▁ beats 21/29 | 60% |
other tests5 not in a task · 38% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Omnilingual ASR paper · FLEURS-81 test mean CER (Whisper-covered languages) lab | 5.6% ✎ | ▅···▃···▃▃██ beats 10/10 | 13% |
| Omnilingual ASR paper · Common Voice 22 (76 languages) test mean CER lab | 6.4% ✎ | ▂▂▂····▂▆··█ beats 10/10 | 9.0% |
| Open ASR Leaderboard – Chilean Spanish · 3-dataset average WER | 14.8% ✎ | ▂▃·····▂▂·█▂ beats 1/8 | 8.4% |
| Open Universal Arabic ASR Leaderboard · 6-test-set average WER | 28.3% ✎ | ▃·▂▂▅█▅▃▂▅▃▂ beats 21/22 | 5.1% |
| Quranic ASR Leaderboard · held-out recitation (600 clips) overall WER | 34.0% ✎ | ▃·▅·▅▃▃·▃▃▃█ beats 1/12 | 3.2% |
$ where to run it
Other ways to run it
- Run it as published · needs ~19.22 GB · from huggingface.co/facebook
uv run --python 3.12 --with omnilingual-asr python -c "from omnilingual_asr.models.inference.pipeline import ASRInferencePipeline as P; print(P(model_card='omniASR_LLM_7B').transcribe(['audio.wav'], batch_size=1))"
Weights and documentation: huggingface.co/facebook ↗
$ how the score was worked out · Open ASR Leaderboard 60%, Omnilingual ASR: Open-Source… 22%, Open ASR Leaderboard – Chilea… 8% 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 |
|---|---|---|
 Open ASR Leaderboard | 1 | 60% |
 Omnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages (arXiv:2511.09690v1) · lab-reported | 2 | 22% |
| Open ASR Leaderboard – Chilean Spanish (idsudd / Alonso Astroza) | 1 | 8% |
| Open Universal Arabic ASR Leaderboard (Elm Research Center) | 1 | 5% |
| Quranic ASR Leaderboard (Quran-Lab, held-out recitation set) | 1 | 3% |
| prior · pull toward the middle | 2% | |
| weighted average | = 57.9 |
$ 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