iOS 27’s AI Model Selection Feature: A New Development in Apple Intelligence

MANDATORY — KEY NEW FACTS

  • In iOS 27, third-party AI models can be selected for each Apple Intelligence feature
  • The “Extensions” mechanism allows AI models to be integrated via the App Store
  • ChatGPT, Gemini, and Claude are available in Siri, Writing Tools, and Image Playground
  • Different Siri voices can be set for each AI model

What’s Changed

In iOS 27, the fundamental architecture of Apple Intelligence is changing. Until now, Apple Intelligence has relied on Apple’s proprietary AI models, but it is evolving into a platform that allows the selection of third-party AI models.

According to Mark Gurman of Bloomberg, in iOS 27, iPadOS 27, and macOS 27, users will be able to set third-party AI services like ChatGPT, Gemini, and Claude as the default for Apple Intelligence features. This change will allow users to use their chosen AI model not only in Siri but also in system-wide AI features like Writing Tools and Image Playground.

Apple is also planning to use a Gemini-based model in iOS 27’s Apple Intelligence and Siri features, thanks to a new contract with Google. (Source: The Verge)

Extensions Integration Mechanism

The new “Extensions” system enables the integration of third-party AI models. AI model providers can opt-in to this system by adding compatibility to their App Store applications.

This Extensions mechanism allows developers to directly integrate their AI models into Apple’s ecosystem. Users can download compatible apps from the App Store and select the corresponding AI model as the default for each Apple Intelligence feature in the settings.

Notably, different Siri voices can be set for each AI model. For example, Siri instances using Apple’s AI model can use the traditional voice, while those using ChatGPT can use a different voice. (Source: Channel News Asia)

Impact on Implementation

This change has significant implications for iOS app developers. Apps that have used the Apple Intelligence API will need to respond to the different characteristics and feature sets of the user-selected AI models.

For example, apps using Writing Tools will need to consider the differences in text generation capabilities between ChatGPT, Gemini, and Claude when designing the user experience. Similarly, apps using Image Playground will need to handle the varying image generation capabilities and limitations of each AI model.

Developers will need to check the detailed technical specifications of the Apple Intelligence API and its compatibility with each AI model when the iOS 27 beta is released. (Source: First Post)

Summary

From this article, developers can:

  • Check the implementation specifications of the Extensions system in the iOS 27 beta and prepare to integrate their AI models into the Apple ecosystem
  • Modify existing apps that use the Apple Intelligence API to support multiple AI models, allowing for flexible response to user selection
  • Develop differentiated features that leverage the characteristics of each AI model (ChatGPT, Gemini, Claude) and provide optimized experiences based on the user’s AI model selection
  • Plan a distribution strategy for AI models on the App Store, expanding their reach to Apple’s vast user base through Extensions compatibility