whatmodel.sh / stt / QuartzNet 15x5 (NVIDIA DLE, 21.07 recipe)
QuartzNet 15x5 (NVIDIA DLE, 21.07 recipe)
↗ share
+ contribute
QuartzNet 15x5 (NVIDIA DLE, 21.07 recipe) is a closed-weights speech-to-text model from NVIDIA, released July 2021.
Catalogue identity notes
NVIDIA DeepLearningExamples training recipe; exact reported A100 run. Do not equate this training result with a differently trained NeMo/NGC checkpoint. Public code and training recipe; exact scored checkpoint not verified.
Score within the LibriSpeech test-clean WER comparison set only.
$ how it compares
Scored speech-to-text models released closest to QuartzNet 15x5 (NVIDIA DLE, 21.07 recipe), out of 176 with a score, then more from NVIDIA.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 144 | STT En Conformer-Transducer Small | 43±3 | local | ||
| 146 |  Vosk en-us 0.22 | 42±6 | local | ||
| 165 |  HuBERT X-Large (LibriSpeech 960h ft) | 16±4 | local | ||
| 166 | HuBERT Large (LibriSpeech 960h ft) | 15±4 | local | ||
| 167 | wav2vec 2.0 Large Robust (LibriSpeech 960h ft) | 15±5 | 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
other tests2 not in a task · 32% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LibriSpeech test-clean WER lab | 3.7% ✎ | ▁·▁·▁▁▁▃▃▇█▆ beats 11/100 | 17% |
| LibriSpeech test-other WER lab | 10.7% ✎ | ▃▁▁▂▅▃▅▄█▂▂▁ beats 0/54 | 16% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · NVIDIA DeepLearningExamples J… 32%
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 DeepLearningExamples Jasper and QuartzNet READMEs · lab-reported | 2 | 32% |
| prior · pull toward the middle | 68% | |
| weighted average | = 35.0 |
$ 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