Apple

Apple Inc. develops iPhone, iPad, Mac, and the software and services around them. Its AI work includes Apple Intelligence, introduced on June 10, 2024, and MLX, an open-source array framework for training and running models on Apple silicon. Apple Intelligence brought generative models into the company's operating-system strategy, combining on-device processing, larger models on Apple silicon servers, and access to OpenAI's ChatGPT.[11][4][7]

The June 2024 introduction placed Apple Intelligence inside iOS 18, iPadOS 18, and macOS Sequoia. Announced features included systemwide Writing Tools to rewrite, proofread, and summarize text; Image Playground for on-device image creation; and natural-language search in Photos. Apple also outlined Siri capabilities that would use personal context and take actions across apps—for example, adding an address received in Messages to a contact card. These were staged plans: Apple announced a free U.S. English beta for fall 2024 on iPhone 15 Pro and Pro Max and M1-or-later iPad and Mac, with some features and languages to follow over the next year.[4]

Private Cloud Compute extends the architecture beyond the device for more complex requests, using larger models on dedicated Apple silicon servers. Apple described safeguards including inspectable server code and a cryptographic requirement that server software be publicly logged before devices communicate with it. Its stated privacy design was that data would never be retained or exposed.[4]

Apple separately announced ChatGPT access through Siri and Writing Tools, powered by OpenAI's GPT-4o. The Siri design asked permission before sending questions and any accompanying documents or photos to ChatGPT. Apple said users could access it without an account, with IP addresses obscured and requests not stored by OpenAI; users who connected an account would instead be subject to ChatGPT's data-use policies.[4]

In research attributed to Apple, Pratyush Maini, Skyler Seto, He Bai, David Grangier, Yizhe Zhang, and Navdeep Jaitly proposed WRAP, or Web Rephrase Augmented Pre-training. Their paper, submitted January 29, 2024, describes using an off-the-shelf instruction-tuned model to paraphrase web documents in styles such as “like Wikipedia” or “question-answer format,” then jointly pretraining on original documents and synthetic rephrases.[1][3]

The WRAP authors reported approximately threefold faster pretraining on C4. At the same pretraining compute budget, they reported more than 10% average perplexity improvement across different subsets of the Pile and more than 2% improvement in zero-shot question-answer accuracy across 13 tasks. They attributed the gains to higher-quality rephrased data and style diversity that more closely reflected downstream evaluation formats.[3]

MLX makes Apple silicon's unified memory architecture part of its programming model: the CPU and GPU access the same memory pool, so developers select a device for an operation rather than move arrays between separate device memories. Its scheduler manages dependencies between CPU and GPU operations. MLX supports neural-network training and inference, with Python and Swift frontends and C and C++ APIs. MLX LM adds language-model generation and fine-tuning; quantization reduces the precision used to store model parameters, lowering their memory footprint.[6][7]

In November 2025, Apple described MLX support for the M5 GPU's Neural Accelerators, which provide dedicated matrix-multiplication operations and require macOS 26.2 or later for their enhanced performance. Comparing M5 and M4 MacBook Pro systems with 24GB of unified memory, Apple reported time-to-first-token speedups of 3.33–4.06 times across six model configurations, but subsequent-token generation speedups of 1.19–1.27 times. The benchmarks used 4,096-token prompts and 128 generated tokens. The distinction reflects two bottlenecks: first-token processing was compute-bound, while subsequent generation was bounded by memory bandwidth.[7]

The devices and chip architecture used by Apple's AI systems grew out of earlier product generations. Apple was founded on April 1, 1976, incorporated in California on January 3, 1977, and held its initial public offering on December 12, 1980. Its product history includes Apple II, Macintosh—introduced in 1984—and later iPod, iPhone, iPad, Apple Watch, and Apple Vision Pro.[10][8][4]

Apple introduced iPhone on January 9, 2007, combining a mobile phone, a widescreen iPod, and Internet communications in a handheld device controlled through a multi-touch display. The announcement named Cingular as its exclusive U.S. carrier, Yahoo as a provider of push IMAP email, and Google as a search and maps provider. Apple planned U.S. availability for June 2007.[9]

On June 22, 2020, Apple announced the Mac's transition to custom Apple silicon, building on more than a decade of its silicon team's work for iPhone, iPad, and Apple Watch. The decision extended a common architecture to the Mac and brought access to technologies such as the Neural Engine for machine learning. Apple planned the first shipments by year-end and a transition lasting about two years.[5]

Apple paired the architecture change with concrete migration tools. Universal 2 binaries supported both Intel and Apple silicon Macs, while Rosetta 2 translated existing applications that had not yet been updated. A $500 developer program supplied documentation, software betas, and temporary use of a Developer Transition Kit: a Mac mini containing an A12Z Bionic chip, 16GB of memory, and a 512GB SSD. Developers had to return the kit at the end of the program.[5]

Tim Cook joined Apple in 1998 and became CEO in 2011. On April 20, 2026, Apple announced a board-approved succession naming John Ternus as the next CEO, effective September 1, and Cook as executive chairman. Ternus had joined Apple's product design team in 2001, become a vice president of Hardware Engineering in 2013, and joined the executive team as senior vice president in 2021. Apple's leadership page identifies Ternus as CEO and a board member.[12][13]

Apple reported $102.5 billion in revenue for its fiscal 2025 fourth quarter, which ended September 27, 2025, up 8% year over year. CFO Kevan Parekh put full-year revenue at a rounded $416 billion in the October 30 results announcement. In April 2026, Apple reported more than 500 retail stores and an active installed base exceeding 2.5 billion devices—a device count rather than a count of unique users.[11][12]