whatmodel.sh / stt / Parakeet RNNT 1.1B
Parakeet RNNT 1.1B
↗ share
+ contribute
Parakeet RNNT 1.1B is an open-weights speech-to-text model from NVIDIA, released January 2024. Its published size is 1.07B parameters. Licence: cc-by-4.0.
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 Parakeet RNNT 1.1B, out of 176 with a score, then more from NVIDIA.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 52 |  Voxtral Mini 3B (2507) | 58±1 | api $0.0010/audio min +local | ||
| 53 |  Universal-Streaming (English) | 58±3 | api $0.0025/audio min | ||
| 54 | Parakeet RNNT 1.1B | 58±2 | local | ||
| 55 |  Chirp 2 | 58±3 | api price unknown | ||
| 56 | Canary-1B-v2 | 58±1 | local | ||
| more from NVIDIA | |||||
| 28 | Canary-Qwen-2.5B | 61±2 | local | ||
| 35 | Parakeet TDT 0.6B v2 | 60±1 | local | ||
| 37 | Parakeet Unified EN 0.6B | 60±4 | local | ||
see all speech-to-text models · find the ones that fit your machine
$ what it's good at
long-form transcription▁▁▄▁·▁▃█▄▁▁▁ #19 of 27 * · 8.3% of its score · 1 test
Trails Whisper large (v1), Whisper large-v2, Rev AI Machine (Reverb ASR) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Open ASR English long-form WER | 10.7% ✎ | ▁▁▅▁·▃▅█▆▁▂▂ beats 9/30 | 8.3% |
other tests13 not in a task · 91% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| AfriSpeech-MultiBench · 7-region African-accent mean WER | 25.6% ✎ | ▂····▂···▂▆█ beats 3/11 | 16% |
| AfriSpeech-MultiBench · Parliament (legal domain) WER | 26.8% ✎ | ▂▂····▂·▅·▃█ beats 5/12 | 13% |
| LibriSpeech test-clean WER +1 | 1.5% ✎ | ▁···▁▁▂·▄▄██ beats 48/56 | 12% |
| SPGISpeech test WER lab | 3.1% ✎ | ▄▄··▄██··▄██ beats 3/11 | 8.5% |
| Open ASR English short-form WER (March 2026) | 7.1% ✎ | ▁▁▁▃··▁▁▂██▆ beats 56/86 | 8.3% |
| Earnings-22 WER (short-form, Open ASR / ESB test split) lab | 14.1% ✎ | █······███·█ beats 0/4 | 7.0% |
| AMI WER (Open ASR / ESB test split) lab | 17.1% ✎ | ▄▄·······█·▄ beats 0/4 | 6.7% |
| GigaSpeech test WER lab | 10.0% ✎ | ▁····▁··▁▄█▂ beats 10/17 | 6.2% |
5 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LibriSpeech test-other WER lab | 2.5% ✎ | ▁▁▁▂▅▃▅▄█▂▂▃ beats 54/54 | 5.0% |
| FLEURS English test WER | 4.5% ✎ | ▁··▁▁▁▁▁▁▃█▁ beats 39/56 | 4.2% |
| VoxPopuli English test WER lab | 5.4% ✎ | █▄▄··████▄██ beats 13/16 | 3.4% |
| TED-LIUM test WER (short-form) lab | 3.9% ✎ | ▁·▁▇█▃·▁▁▃█▄ beats 19/27 | 0.3% |
| Common Voice 9.0 English test WER lab | 5.8% ✎ | no other model yet | — |
$ where to run it
Other ways to run it
- Run it as published · needs ~5.64 GB, NVIDIA only · from huggingface.co/nvidia (official)
uvx --with 'nemo_toolkit[asr]' python -c "import nemo.collections.asr as a; print(a.models.ASRModel.from_pretrained('nvidia/parakeet-rnnt-1.1b').transcribe(['audio.wav']))"
Official weights and documentation: huggingface.co/nvidia ↗
$ how the score was worked out · NVIDIA ASR model cards (Parak… 39%, AfriSpeech-MultiBench (arXiv:… 29%, Open ASR Leaderboard 17% and 1 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 |
|---|---|---|
 NVIDIA ASR model cards (Parakeet / Canary, Hugging Face) · lab-reported | 8 | 39% |
 AfriSpeech-MultiBench (arXiv:2511.14255v1) · African-accented English ASR benchmark suite | 2 | 29% |
| Open ASR Leaderboard | 2 | 17% |
 transcribe.cpp model catalog · LibriSpeech test-clean WER of validated full-precision ports | 2 | 15% |
| prior · pull toward the middle | 1% | |
| weighted average | = 58.0 |
not counted (1)
| NVIDIA ASR model cards (Parakeet / Canary, Hugging Face) | Common Voice 9.0 English test WER [correct] | 5.8% | fewer than three models have this test yet, or this source ran it for this model alone |
$ 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