For years, asking Apple to welcome a rival's AI into its own tools felt like asking a cat to fetch your slippers. This week the cat didn't just fetch the slippers — it fetched Claude's, Gemini's, and OpenAI's, then offered to wire up a multi-agent workflow while it was at it.

One Swift API, Many Brains

At its 2026 developer announcements, Apple expanded the Foundation Models framework into a single native Swift API supporting on-device models, image input, and — the headline act — third-party models. Developers can now call Claude, Gemini, "or those from any other provider" through the same interface, with Apple specifically highlighting a custom Gemini integration built alongside Google.

Apple also introduced a new Core AI framework for running full-scale LLMs locally on Apple silicon's Neural Engine, plus an updated App Intents framework that plugs apps into Siri's onscreen awareness and personal context. Xcode 27 now bundles coding agents from Anthropic, Google, and OpenAI, adds a Device Hub for managing hardware and simulators, ships 30% smaller, and runs Xcode Cloud builds up to twice as fast.

The Walled Garden Grows a Side Gate

The genuinely surprising part is the price tag: developers in the App Store Small Business Program with fewer than 2 million first-time downloads get cost-free access to Apple's next-generation Foundation Models running on Private Cloud Compute. "Free server-grade inference" is a phrase Apple almost never lets out of the building.

Squint, and the strategy snaps into focus. Apple doesn't necessarily need to win the model race outright if it owns the device, the API, and the distribution. Let Claude and Gemini slug it out — Apple would quite happily be the toll booth every one of them has to drive through.

The walled garden is still very much walled. It just installed a remarkably convenient side gate — and kept every copy of the key.

Source: Apple Newsroom