Apple Intelligence is the term under which Apple grouped all the generative AI features coming later this year on iPhones, iPads, and Macs. Users will be able to use Apple Intelligence features to reply to emails quickly, summarize the content of websites or messages, reorganize and prioritize notifications, create images from text, and much more. The system uses a series of Apple-developed LLMs to satisfy customer requests, leveraging its ability to control the devices' software and hardware.

Apple says Apple Intelligence is powerful, intuitive, integrated, personal, and private. To enable this "personal" intelligence, Apple could leverage its standing on privacy and the trust of its customers. The system will be capable of reading screens and indexing data on the phone, even private information, with no risk to the user's privacy, as most of the computing happens on the device and stays there, inaccessible to anyone but the user.

Most of Apple Intelligence's features will be enabled by foundation models capable of running inference on the device. For operations that need a more powerful model, Apple devised a solution called Private Cloud Computer, a new generation stateless server with no logs or memory that can handle remote compute without ever compromising the privacy of the data transferred to the cloud.

The local nature of most of Apple's Intelligence Models, plus the development of Private Cloud computing, made it possible for Apple to hop on the generative AI train without compromising its fundamental stance on privacy, a position that many other relevant competitors have been willingly overlooked in the name of AI innovation.

Apple's new models also enable a completely new Siri that's more intelligent and capable thanks to full context awareness. Siri can reply to contextual questions about information in the users' ecosystem. Users will be able to ask specific questions about personal matters, and Siri will be capable of answering thanks to the knowledge it acquired by indexing the device's content, from mail to images to messages.

And for more general-knowledge questions, Apple's assistant will be able to tap into generalistic, external models, starting with chatGPT 4o by OpenAI.

Apple's choice of mentioning chatGPT from the start has created some confusion, with many misunderstanding the role of OpenAI's model. ChatGPT will be integrated into the systems modularly and will not be a fundamental part of Apple Intelligence. Think of it as an external module that users can seamlessly tap into when they need to work with the kind of general knowledge context that the model can provide.

Apple Intelligence will be available in iOS 18, iPadOS 18, and macOS 15 Sequoia, but only on Macs and iPads with an M1 chip or higher and iPhones with an A17 Pro chip or higher. This limits the selection to iPhone 15 Pro and Pro Max, plus probably the entire iPhone 16 lineup expected to arrive in the fall.