The best MacWhisper model for your Mac: WhisperKit and Turbo on Apple silicon, cloud on Intel
By Ryleigh Newman · Published
On an Apple silicon Mac, pick a WhisperKit model in MacWhisper. Its help center says WhisperKit gives better results, especially on long files with silence or music, and runs faster, and it calls the Turbo model a great choice for many Macs. On an Intel Mac, WhisperKit is not available, and MacWhisper's own advice for speed is one of its cloud transcription providers.
I built YapToText, a dictation app for the Mac, so read what I say about MacWhisper with that in mind. Every fact about MacWhisper here comes from its own help center, read on October 11, 2026.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
Pick by the Mac you have
From MacWhisper's help articles on WhisperKit and on slow transcription, read October 11, 2026.
| Your Mac | What MacWhisper's help center suggests |
|---|---|
| M-series, most models | A WhisperKit model. Turbo is its suggestion for many Macs |
| M-series Pro, Max or Ultra | The same, plus the speed settings below |
| Intel | WhisperKit is not supported. For speed, a cloud transcription provider |
Switch to a WhisperKit model
- Open the Manage Models screen from the model menu.
- Filter the list by WhisperKit.
- Download a model. Turbo is the one the article names.
- Wait for the first load.
The first time a model is used it is prepared for your Mac, which can take up to a few minutes for the larger models. Later loads are fast. WhisperKit models run only on M-series Macs.
What the model choice fixes
- A transcript that ends in repeated words. MacWhisper's article says to switch to a WhisperKit model, which does not have that issue. Whisper hallucinations explains why the older models loop.
- Speaker names. Only WhisperKit models, and two of the cloud providers, label speakers. In the model list the local ones carry a green Speaker Recognition badge. The speaker recognition page has the steps.
- Long files with silence or music. The case the help center singles out as better with WhisperKit.
If transcription is slow
- On an Intel Mac. Intel Macs lack dedicated machine learning hardware, so local transcription is much slower. The help center suggests a cloud transcription provider.
- On an M-series Mac. MacWhisper may have crashed earlier and dropped into a compatibility mode, which the article says is common on M1 Macs. Open Settings, Advanced, WhisperKit and make sure Disable all GPU usage is off.
- On a Pro, Max or Ultra chip. For maximum speed, the same screen lets you enable VAD, set Concurrent Workers to 64 and set the Text Decoder to GPU.
Cloud transcription and its file limits
Cloud transcription sends the file to a provider you add in Settings, on the cloud transcription page, with your own API key. You then choose that provider from the model picker in the top right of the home screen.
From MacWhisper's cloud transcription article, read October 11, 2026. It adds that 1 GB uploads may not work reliably on a slow connection.
| Provider | File size limit |
|---|---|
| ElevenLabs | 1 GB |
| Deepgram | 1 GB |
| Two other providers it names | 25 MB each |
MacWhisper also sells its own cloud transcription through the optional Assistant subscription, and its article says not to use that with sensitive data, because the file goes to its servers. The MacWhisper privacy page covers what leaves the Mac.
One model, already chosen
MacWhisper gives you a model list because it serves many Macs and many jobs. YapToText, the free app I built, makes the choice for you on Apple silicon: Whisper Large v3 Turbo ships inside the App Store download, so there is nothing to pick or prepare before the first dictation. It has a model library if you want another one, and three English-only models. It does not run on Intel Macs and has no cloud option. Parakeet vs Whisper compares the two model families.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
The short version
- Apple silicon. A WhisperKit model, with Turbo as the starting point.
- Intel. A cloud provider, for speed.
- Repeated words. Switch to WhisperKit.
- Slow on M-series. Turn off Disable all GPU usage.
- Cloud limits. 25 MB or 1 GB, by provider.
If MacWhisper changes its models, open an issue on GitHub and I will update this page. Every one gets read.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
Related
Sources
- https://docs.macwhisper.com/article/29-switching-to-a-whisperkit-model (checked 2026-10-11)
- https://docs.macwhisper.com/article/34-transcription-is-slower-than-expected (checked 2026-10-11)
- https://docs.macwhisper.com/article/20-what-to-do-with-repeated-text (checked 2026-10-11)
- https://docs.macwhisper.com/article/18-cloud-transcription (checked 2026-10-11)
- https://docs.macwhisper.com/article/22-assistant (checked 2026-10-11)
- https://github.com/ryleighnewman/YapToText (checked 2026-10-11)
- https://apps.apple.com/us/app/yaptotext/id6786382289 (checked 2026-10-11)
Questions
What is the best model in MacWhisper?
Its help center recommends WhisperKit models on M-series Macs and calls Turbo a great choice for many Macs.
Does WhisperKit work on Intel Macs in MacWhisper?
No. MacWhisper's article says WhisperKit models are only available on M-series Macs.
Why is MacWhisper slow on my M1?
It may be in a compatibility mode after an earlier crash. In Settings, Advanced, WhisperKit, make sure Disable all GPU usage is off.
Why does MacWhisper repeat words at the end of a transcript?
An older model is looping. MacWhisper's help center says to switch to a WhisperKit model.
What is the file size limit for MacWhisper cloud transcription?
It depends on the provider: 25 MB for two of them, and 1 GB for ElevenLabs and Deepgram.