Question · Dictation software
Which MacWhisper model is best?
For most people on an Apple silicon Mac, Whisper Large v3 Turbo is the best default model in MacWhisper. Choose Parakeet v3 (Pro) for noisy audio, accents or hours of recordings, and a smaller Whisper model on an Intel Mac or one with 8 GB of memory.
Updated October 27, 20263 min read21 cited facts, each traced to its source
Whisper Large v3 Turbo is the best default model in MacWhisper on a Mac with an M-series chip. It is the most accurate of the common options on clean English, and it supports the widest range of languages. Parakeet v3 is the better pick when the audio is noisy, the speakers have varied accents, or the file is hours long. These conclusions come from our own tests of the same engines MacWhisper runs, on a Mac with an M2 Pro with nothing sent over the network. The full model discussion is in our MacWhisper guide.
Turbo wins on clean English
We measure accuracy as word error rate, the share of words wrong. On 100 clean English recordings, Whisper Large v3 Turbo made 3.92 percent errors 1, Parakeet TDT v3 made 4.48 percent 2 and Apple's on-device model 6.73 percent 3. The gap between the first two is small.
Turbo is OpenAI's slimmed-down Large v3, with 4 decoder layers instead of 32 4, and it runs about 8 times faster than the large model 5. MacWhisper reaches over 100 languages 6, while Parakeet v3 covers 25 European languages 7. Outside Europe, Whisper is the choice.
Parakeet wins when audio is hard
On English conversations from 13 accent groups, Parakeet's error rate was 12.57 percent, against 17.66 percent for Whisper 8 9. With talk in the background 5 dB quieter than the speaker, Whisper's errors rose to 13.92 percent 10 while Parakeet's rose to 7.85 percent 11.
Long files favor Parakeet as well. Whisper is trained on 30-second chunks of audio 12, and in our test it dropped a passage of 10 or more words in 4 of 84 whole-file readings, against 1 of 84 for Parakeet 13 14. Parakeet is also faster: 9to5Mac transcribed a 3-hour podcast with speaker recognition in 1 minute 22 seconds in MacWhisper 15.
It depends on your Mac, license and task
| Your situation | Pick |
|---|---|
| Clean English, or a language outside Europe | Whisper Large v3 Turbo |
| Meetings, noise, accents, long recordings | Parakeet v3 (Pro, M-series Mac) |
| Speaker labels | A model with the Speaker Recognition badge (see speaker recognition) |
| Translating speech into English | Whisper Medium or Large, since Turbo is not trained for translation 16 |
| Intel Mac or 8 GB of memory | A smaller Whisper model, or a cloud provider |
Three limits to know. Parakeet is a Pro feature, and when MacWhisper first added it, it was English only 17, so check the current model list. Medium and Large need a Mac with more than 8 GB of memory 18. And Intel Macs lack dedicated machine learning hardware, which makes on-device transcription significantly slower 19.
What to do
- Start with Large v3 Turbo. The free version supports Whisper models, though some are only fully available in Pro 20. See whether MacWhisper is free.
- Switch to the WhisperKit version of the model if you have an M-series Mac and long files with silence or music, as the maker says it gives better results there 21.
- Test Parakeet v3 on your own worst recording. If errors fall, keep it for meetings. For recording calls see meeting recording.
For the same question about Whisper in general, see which Whisper model is best, and for the numbers behind accuracy, Whisper vs Parakeet accuracy. Our overview of dictation software places MacWhisper among its rivals.
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.
- 13.
- 14.
- 15.
- 16.
- 17.
- 18.
- 19.
- 20.
- 21.
The data behind these numbers
- On-Device Speech Recognition Benchmark: 25 Languages3 of this page's figures, 56 data points in all
- Speech Recognition Accuracy: Word Error Rates3 of this page's figures, 864 data points in all
- Speech Recognition and Accents: 13 English Accents Tested2 of this page's figures, 42 data points in all
- Dictation in Noise: Speech Recognition Accuracy Tested2 of this page's figures, 15 data points in all
- How Many Languages Speech Recognition Supports1 of this page's figures, 414 data points in all
Go deeper
Related questions
- Can MacWhisper record and transcribe meetings?
- Does MacWhisper recognise different speakers?
- Is MacWhisper free?
- Aqua Voice vs Wispr Flow: what is the difference?
- Can you run voice recognition software on a Raspberry Pi?
- Can you upgrade an old Dragon NaturallySpeaking to version 15 or 16?
- Does Apple have a voice to text app?
- Does Aqua Voice work on Windows and iPhone?