Apple took the wraps off its new AFM 3 Core Advanced AI model during WWDC26 on Monday, marking a significant push into on-device artificial intelligence. The model packs 20 billion parameters and is designed to run directly on Apple hardware, sidestepping the cloud for many tasks.
What the AFM 3 Core Model Brings
The AFM 3 Core is Apple's latest foundation model, built for tasks like text generation, summarization, and image analysis. With 20 billion parameters, it's larger than previous on-device models from the company, but still small enough to fit on a phone or laptop. Apple says the model can handle complex requests without sending data to remote servers, a move that strengthens its privacy pitch.
Why On-Device AI Matters
Running AI locally cuts latency and keeps user data on the device. For Apple, that's a competitive advantage. Competitors like Google and Microsoft rely heavily on cloud-based AI, which requires an internet connection and raises privacy questions. Apple's approach means features like smart replies, photo editing, and voice assistants could work faster and without network dependence. The company didn't say which devices will ship with the AFM 3 Core, but analysts expect upcoming iPhones and Macs to include the needed neural engine upgrades.
WWDC26 Context
The announcement came during the keynote of Apple's annual Worldwide Developers Conference. The company also previewed updates to iOS, macOS, and watchOS, but the AI news dominated the session. Apple has been investing heavily in generative AI after a slower start compared to rivals. The AFM 3 Core is its most ambitious on-device effort yet.
Apple hasn't set a firm release date for the AFM 3 Core. Developers can access it through the new CoreAI framework in the latest betas. The company is expected to roll out specific features tied to the model in the fall, alongside new hardware. Whether the 20-billion-parameter model will be enough to match cloud-based giants remains an open question.
