Apple iOS 27 to Let Users Pick Their Own AI Model with ‘Extensions’
2026年5月6日

Apple is reportedly preparing a major shift in how AI runs on iPhone. According to a Bloomberg report, the company plans to let users choose third-party AI models for several “Apple Intelligence” features in iOS 27.
This could reshape how generative AI is delivered on smartphones.
What “Extensions” Will Do
The new feature, internally code-named “Extensions,” will let users access generative AI capabilities from installed apps through Apple Intelligence features such as Siri, Writing Tools, and Image Playground.
According to Bloomberg, Google and Anthropic models are already in testing. ChatGPT, currently the default option, is expected to remain as a choice. The same mechanism is set to extend to iPadOS 27 and macOS 27.
Apple’s AI Strategy at an Inflection Point
The shift comes as long-time CEO (chief executive officer) Tim Cook prepares to step down, with John Ternus tapped as his successor and tasked with charting Apple’s AI direction.
Apple has often been described as “behind on AI,” with fewer in-house generative AI services than competitors. But the company has a clear strategy: instead of pouring billions into its own AI infrastructure, it is turning existing hardware into a gateway to AI experiences.
A Platform That Aggregates AI
By offering access to multiple models rather than locking everyone into one, Apple is positioning itself as a platform that aggregates third-party AI rather than building everything in-house. The AI layer of the operating system is becoming something users can choose, much like a web browser.
This signals a shift from “AI bundling” to “AI selection”—a different role for the platform owner.
What App Developers Should Prepare For
For app developers, the change brings three immediate considerations.
First, app design needs to assume model choice. A feature triggered through Siri may run on Google Gemini, Anthropic Claude, or ChatGPT depending on user settings—and absorbing differences in style and accuracy becomes a new UX (user experience) challenge.
Second, API (application programming interface) pricing leverage may improve. With multiple model choices on iOS, dependence on any single vendor decreases, and choosing the optimal model per feature becomes more practical.
Third, prompt design needs standardization. The same feature may produce different output depending on the underlying model, so “model-agnostic” prompt design will likely become a useful corporate guideline.
Key Takeaways
・Apple is preparing an “Extensions” feature in iOS 27 that lets users pick which AI model powers Apple Intelligence functions.
・Google and Anthropic models are already in testing, while ChatGPT is expected to remain as an option.
・As the OS-level AI layer becomes user-selectable, app developers will need to design for multi-model environments.
Sources
TechCrunch (iOS 27 multi-AI feature)


