Prepare local AI on your Mac
Local AI requires a model running on your Mac. Prepare the model and components while connected before relying on offline use. Then test a small file containing no personal information with the network disconnected.
Open the interactive product view ↗If a ready model feels slow
Compare memory pressure and workload in Stats. Try a smaller document and reduce other heavy work running at the same time. CPU percentage alone does not tell you whether a model is working.
Why test offline?
A successful operation while connected does not by itself prove offline operation. Repeating the same task on the same file with the network disconnected helps verify the local workflow you need.
How to use it
- Choose the task: image understanding, document text or naming suggestions. Different tasks may require different models and hardware resources.
- Check model status in the app and complete required downloads. Leave sufficient free disk space; model size and memory needs depend on your choice.
- Try the supported local task with a sample file while online. Confirm the model is ready and produces a useful result.
- Disconnect from the internet and repeat the same local task. Compare the result and time; keep downloads and web archiving separate from this test.
Scope and considerations
There is no fixed RAM, disk or speed guarantee here; requirements depend on model and task. Initial model downloads use the internet. Local file processing does not make separate network operations such as updates or support forms offline.