News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
imgfnApple Intelligence

Apple Intelligence

Apple's on-device and cloud AI features built into iOS, iPadOS and macOS

Apple Intelligence is a set of artificial intelligence features built by Apple into iOS 18, iPadOS 18 and macOS Sequoia, announced on 10 June 2024 and released on 28 October 2024. It runs partly on the device and partly on Apple's Private Cloud Compute servers, and underpins writing tools, image generation, summaries and an updated Siri.

Share on WhatsApp

  • Announced 10 June 2024; released 28 October 2024
  • On-device model has 3 billion parameters; 20B version on newer devices
  • Private Cloud Compute uses custom silicon and verifiable software
  • ChatGPT integration is opt-in and free without signing in
  • Google Gemini partnership announced January 2026

What it is

Apple Intelligence is a collection of artificial intelligence features that Apple builds directly into its operating systems. It was announced on 10 June 2024 at the Worldwide Developers Conference alongside iOS 18, iPadOS 18 and macOS Sequoia, and reached users on 28 October 2024. Work is split between the device itself and Apple's own servers. The features cover writing help, image creation, notification summaries, photo editing and a rebuilt Siri. It is free for anyone with a supported device, which means Apple silicon Macs, or iPhones and iPads with the A17 Pro chip or newer.

How it came about

Apple's AI work began with Siri in 2011, and the company grew its research team from 2015 while buying AI firms through 2020. The arrival of ChatGPT in late 2022 pushed Apple to refocus, and reports in July 2023 described an internal model codenamed "Ajax". Tim Cook promised AI features for 2024. In January 2026 Apple and Google announced a multi-year partnership around Google's Gemini models and cloud, and at WWDC 2026 Apple showed a new version of Apple Intelligence with a rebuilt Siri called Siri AI.

Inside the model

Apple Foundation Models come in two flavours: one runs on the device, the other on servers built around Apple silicon. Each pairs a general foundation model with smaller adapters tuned for jobs such as summarising text or changing tone. The on-device model has 3 billion parameters, with a 20 billion version, AFM 3 Core Advanced, on newer hardware. Cloud work sits on Private Cloud Compute, using custom silicon for end-to-end encryption, with software that researchers can independently verify; devices refuse connections if the code does not match. Apple says these data centres have run on 100% renewable energy since 2018.

What it can do

Writing tools proofread, rewrite and summarise text, and can produce tables and lists. Image Playground and Genmoji generate pictures on-device, while Clean Up removes unwanted objects from photos. Mail sorts messages and highlights urgent ones, Photos builds memory movies from a described story, and notification summaries reduce clutter. Real-time translation covers messages, calls and AirPods conversations. In Apple's own human evaluation, the on-device model beat or tied comparable small models from Mistral AI, Microsoft and Google, and the server models beat GPT-3 while roughly matching GPT-4.

How to use it

Apple Intelligence ships free inside iOS 18.1, iPadOS 18.1, macOS 15.1 and, from 31 March 2025, visionOS 2.4. It needs Apple silicon on Mac, and an A17 Pro or newer chip on iPhone and iPad; Apple says older Neural Engines are not up to the task, and analyst Ming-Chi Kuo points to RAM limits. From June 2025 developers can reach the models through a Foundation Models API with structured output and tool calling. ChatGPT access is opt-in and free without signing in, with paid subscribers getting more GPT-4o requests.

Safety and testing

Private Cloud Compute is designed so that the code running on Apple's servers matches the version researchers can inspect, and devices will not connect if there is a mismatch. Requests sent to OpenAI hide the user's IP address, and users are prompted before any data or photos leave the device. Apple states its data centres have used 100% renewable energy since 2018 and has a goal of carbon neutrality across products and supply chain by 2030.

Reception and debate

Apple's Siri investments were criticised by reviewers and inside the company for falling behind rival assistants, and the ChatGPT launch reportedly caught Apple executives off guard. Features promised at WWDC 2024, notably Siri using personal context, slipped repeatedly and only arrived in beta in late 2026. Analysts noted that memory demands kept older iPhones off the supported list, and the feature cannot run on Macs booted from an external drive.

Why it matters

Apple Intelligence brought generative AI into a mainstream operating system used across iPhone, iPad and Mac, and pushed privacy-focused server design into the debate through Private Cloud Compute. Its Foundation Models opened to third-party developers in 2025. The January 2026 Google partnership, and the later use of Gemini outputs to refine Apple's own models, linked Apple's on-device stack to a trillion-parameter frontier model while keeping processing on Apple devices and servers.

Questions people ask about Apple Intelligence

What is Apple Intelligence?

It is a set of AI features Apple builds into iOS, iPadOS and macOS, covering writing help, image generation, photo editing, summaries and Siri. Some work happens on the device and some on Apple's own servers, and it is free for anyone with a supported device.

Which devices support it?

Macs need Apple silicon, and iPhones and iPads need the A17 Pro chip or newer. Apple says older Neural Engines cannot handle the work, and analyst Ming-Chi Kuo notes RAM limits on older iPhones. Apple Vision Pro gained support on 31 March 2025 with visionOS 2.4.

How does it protect privacy?

Cloud processing runs on Private Cloud Compute, built with custom silicon for end-to-end encryption. The software on those servers is independently verifiable, and devices refuse to connect if it does not match. ChatGPT requests hide the user's IP address and require a prompt before data is sent.

What models are behind it?

Apple Foundation Models come as on-device and cloud versions, each with adapters for specific tasks. The on-device model has 3 billion parameters, with a 20 billion variant on newer hardware. From 2026, Apple's models were refined using outputs from Google's Gemini frontier models.

Does it use ChatGPT?

Yes. Siri can hand complex requests to ChatGPT, powered by GPT-4o, and the integration is opt-in. It is free without signing in, with a limited number of GPT-4o requests before falling back to a weaker model; paid subscribers get more.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "Apple Intelligence", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryLLM rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.