Results
Parakeet V2 accuracy and speed
Parakeet V2 is the previous generation of NVIDIA's Neural Engine model. V3 replaced it in the app, and it is here so you can see what the upgrade bought.
74
Blended score
10.4%
Word error rate
0.06 s
Wait for the text
Macro average over 8 datasets on Apple M4
On Apple M4
| Dataset | Score ↑ | WER ↓ | Speed ↑ | Recall ↑ | F-score ↑ | Response ↓ |
|---|---|---|---|---|---|---|
| AMI Meetingssw-app-local-sw-app-parakeet-v2-esb-ami-20260806 | 76 | 9.8% | 189× | — | — | 0.06 s |
| Common Voice, Australian Englishsw-app-local-sw-app-parakeet-v2-common-voice-24-en-au-20260806 | 85 | 6.1% | 84× | — | — | 0.06 s |
| Common Voice, spontaneous Englishsw-app-local-sw-app-parakeet-v2-common-voice-spontaneous-4-en-20260806 | 60 | 15.8% | 128× | — | — | 0.06 s |
| Earnings 22sw-app-local-sw-app-parakeet-v2-esb-earnings22-20260806 | 72 | 11.1% | 165× | — | — | 0.06 s |
| Earnings 22 (with vocabulary)sw-app-local-sw-app-parakeet-v2-contextual-earnings22-20260806 | 62 | 15.3% | 151× | 90.4% | 91.9% | 0.06 s |
| LibriSpeech Othersw-app-local-sw-app-parakeet-v2-esb-librispeech-other-20260806 | 90 | 4.1% | 165× | — | — | 0.06 s |
| Loquacioussw-app-local-sw-app-parakeet-v2-loquacious-test-20260806 | 82 | 7.2% | 136× | — | — | 0.06 s |
| Phonetic vocabularysw-app-local-sw-app-parakeet-v2-phonetic-vocab-20260806 | 66 | 13.6% | 47× | 71.4% | 76.9% | 0.06 s |
Updated August 27, 2026 from bench commit 5344cab. The second line of each row is the run that produced it.
Method
How this was measured
Local results come from driving the real Superwhisper app over each clip, so they reflect the engine and decode settings that ship in the product rather than a separate implementation.
Word error rate counts insertions, deletions and substitutions against a human transcript, so lower is better. Speed is a multiple of real time, so higher is better. The full method and every other model is on the benchmarks page.
Compare
Other models in the benchmark
Same audio, same machines, same method.
Support