Apple Intelligence is the artificial intelligence system Apple has built into the iPhone, iPad, and Mac — a set of features that rewrite and summarize text, generate images and custom emoji, and give Siri a better grasp of what’s on your screen and in your apps. It runs on a mix of small models processed entirely on the device and larger models that run on Apple’s own private servers, an architecture built specifically to avoid routing routine requests through a generic cloud.

What Apple Intelligence actually is

Apple Intelligence launched on October 28, 2024, with iOS 18.1, iPadOS 18.1, and macOS Sequoia 15.1. It isn’t a separate app or a subscription — it’s a layer of AI features built into the operating system itself, free for anyone with a compatible device.

The feature set includes:

  • Writing Tools — proofreading, rewriting, and summarizing text across Mail, Notes, Messages, and most other apps.
  • Genmoji and Image Playground — generating custom emoji and images from a written description.
  • Notification and message summaries — condensing long email threads or group chats into a line or two.
  • A revamped Siri — able to draw on personal context, such as files, messages, and photos, to answer more specific questions and act across apps.
  • ChatGPT integration — an opt-in bridge that hands certain requests to ChatGPT with the user’s permission, without requiring an OpenAI account.

Because the on-device half of the system depends on the device’s own chip, Apple Intelligence only runs on newer hardware: iPhone 15 Pro or later, and any iPad or Mac built on an M1 chip or newer. Older devices can install the same iOS or macOS update but won’t get the AI features.

How it works: two tiers of processing

Apple splits requests between two places. Simple tasks — proofreading a paragraph, summarizing a notification — run on a compact large language model stored directly on the device, small enough (around 3 billion parameters) to sit in memory alongside everything else the phone is doing. None of that data leaves the device.

When a request needs more reasoning power than the on-device model can supply, Apple Intelligence can route it to Private Cloud Compute (PCC) — servers built on Apple’s own chips, running a stripped-down operating system that Apple says even its own staff cannot access. Apple states that PCC discards a user’s data as soon as it has answered the request, and it publishes the server software so independent researchers can verify the servers run only the code Apple claims they run. That verifiability is the main way this setup differs from a typical cloud AI service, where users generally have to take a provider’s privacy claims on faith.

Why it matters

Apple Intelligence is a bet that mainstream AI features — the kind hundreds of millions of people now expect on a phone — don’t require handing every request to a distant, opaque server. With more than 2.5 billion active Apple devices in daily use worldwide, the privacy architecture the company picked effectively becomes a baseline the rest of the industry gets measured against.

It also exposes the limits of building AI alone. Apple’s own on-device and cloud models remain modest next to frontier systems from OpenAI, Google, or Anthropic, which is why Apple leans on outside partners for its hardest requests — ChatGPT in the US and Europe, and, since mid-2026, Alibaba’s Qwen models for mainland China, where Apple’s own cloud servers cannot legally operate.

In the news

China withheld approval for Apple Intelligence for more than two years, since the system’s cloud tier didn’t run through servers registered inside the country, and Chinese law requires locally approved generative-AI services. Apple resolved that by rebuilding the mainland version around Alibaba’s Qwen models for language tasks and Baidu’s technology for search and voice — see our brief on China approving Apple Intelligence, powered by Alibaba’s Qwen. Outside China, Apple’s own on-device and cloud models continue to handle the same features.

FAQ

Is Apple Intelligence free?
Yes. It ships with a qualifying device’s OS update at no extra cost, with no subscription required. Apple’s support guide covers how to check compatibility and turn it on.

Which iPhones support it?
iPhone 15 Pro and Pro Max, and every iPhone 16 model and newer. Older iPhones can update their software but won’t get the AI features, since they lack the required chip.

Does Apple see my requests?
Apple says on-device requests never leave the phone. For requests sent to Private Cloud Compute, Apple says the data is used only to answer that request and deleted immediately afterward — not stored, and not used to train models.

Is it the same everywhere in the world?
No. Feature availability and the underlying models differ by region: the EU saw a delayed rollout of newer Siri features tied to regulatory review, and mainland China runs on Alibaba’s and Baidu’s models rather than Apple’s own cloud, under local law.

Sources: Apple Newsroom, Apple Security Research (Private Cloud Compute documentation), and Wikipedia.