Apple Intelligence vs local AI: what's the difference?

Apple Intelligence is the AI Apple built into macOS and iOS — writing tools, summaries, a smarter Siri, baked into the apps you already use. A local-AI app is one you install on top, like Outlier. The honest difference is three things: who owns the model, whether every request stays on your machine, and how big a model you can run. Apple Intelligence is convenient and free; a local-AI app trades that convenience for control.

What each one is

Apple Intelligence is Apple's set of AI features built into macOS and iOS. It powers Writing Tools (rewrite, proofread, summarize), notification and mail summaries, image features, and a more capable Siri. It runs on Apple Silicon, it's free with a recent OS, and there's nothing to download or set up. Architecturally, Apple Intelligence runs lighter tasks on-device with Apple's own compact model and routes heavier requests to Apple's Private Cloud Compute servers — Apple-run hardware designed so your data isn't retained. The model is Apple's, and you use what Apple exposes through the system.

A local-AI app is a separate program you install — Outlier is one example. You pick the model, the app loads it onto your Mac, and it runs there. Because the whole model lives on the machine, it can work with Wi-Fi switched off. You're not limited to one vendor's model or one menu of features; instead of OS-level writing tools, an app like Outlier gives you a full chat workspace plus a coding agent, deep research, and vision. The trade is that you install and manage it yourself rather than it just being there.

The honest contrast

This isn't a "which is smarter" question — it's a "which job" question. Apple Intelligence is built for the quick, in-context stuff: tidy up this email, summarize this thread, ask Siri to do a thing. It's right there, it's free, and most people will never need to think about it. That convenience is real.

Where a local-AI app pulls ahead is control and capability. With Apple Intelligence, the model is Apple's, the feature set is what Apple ships, and some requests leave the device for Private Cloud Compute. With a dedicated app, you pick the model, every request can stay on the Mac, and you can run models in a weight class Apple's compact on-device model doesn't reach. Owning the model instead of renting access to one is the whole point.

Apple Intelligence vs local AI — the comparison

Dimension	Apple Intelligence	Local-AI app (e.g. Outlier)
Who makes the model	Apple	You choose (open-weight models; import your own MLX model)
Fully on-device?	Some tasks on-device; heavier ones go to Private Cloud Compute	Yes — every request stays on the Mac
Works with Wi-Fi off?	Only the on-device tasks	Yes — fully offline
Model size / power	Apple's compact on-device + cloud models	Up to a 397B-parameter model on a 64 GB Mac (paged inference)
Openness	Closed — Apple's model and feature set	Open-weight models, published on HuggingFace
What it does	Writing tools, summaries, Siri, OS-level features	Chat, coding agent, deep research, vision, computer use
Price	Free (built into a recent macOS)	Free Nano + Lite; Pro $20/mo or $149/yr; lifetime from $99

So which one should you use?

For most people the answer is both — they barely overlap. Keep Apple Intelligence on for the in-line writing help and Siri; it's free and already there. Reach for a local-AI app when the job is bigger than "rewrite this paragraph": you want a specific model, you need the work to never leave your Mac (legal drafts, client files, source code), or you want a coding agent and research Apple doesn't put in the OS.

The privacy line matters most for sensitive work. Apple designed Apple Intelligence carefully, but its heavier requests still travel to Apple's servers. A local app that runs entirely on-device means the question never leaves the room — which is exactly why people keep code and confidential work on a local model. Files on your disk don't have a terms-of-service page.

Frequently asked questions