In short
Running a language model yourself means loading gigabytes of weights into memory, which hurts on 8 GB. Apple Intelligence already includes a model that macOS manages for every app. Kempt uses that one, so there is nothing extra to download, and Kempt’s own process peaked at about 280 MB while naming 25 files in our test.
Without any app
Check your memory pressure
Open Activity Monitor › Memory. Green pressure means you have room; yellow or red while an app runs means it is too heavy for your Mac.
Prefer apps that use Apple’s model
An app built on Apple Intelligence shares the model macOS already keeps. Apps that bring their own model add that model’s size to your memory and disk.
Let big jobs run while you’re away
Naming hundreds of files is work for any model. Start it before a break rather than during a video call.
With Kempt
No model to download
Kempt reads files with the model built into macOS 26. Turn on Apple Intelligence once and Kempt works offline.
Measured, not promised
Naming 25 real downloads, Kempt’s own peak memory was about 280 MB. Searching 1,600 documents takes around 10 ms.
Steps back when you need the Mac
Background naming pauses on low power or heat and yields to what you’re doing, then resumes.
$27 once, 14 days freeRuns on every Apple silicon MacA question? Ask a person
Questions people ask
Does Apple Intelligence work on an 8 GB Mac?
Yes, on Macs with Apple silicon (M1 or later). macOS downloads and manages the model itself, separately from any app.
Will Kempt slow down my MacBook Air?
The first pass over a full Downloads folder keeps the model busy for a few minutes. After that, Kempt only names new files as they arrive.
Does it work on an Intel Mac?
No. Apple Intelligence needs Apple silicon, and so does Kempt.