Apple unveiled a significantly upgraded version of Siri AI at WWDC 2026, moving beyond previous promises that many developers and consumers viewed with skepticism. The update relies on a custom model derived from Google’s Gemini architecture, which runs on Apple’s own private cloud compute infrastructure rather than public servers. A key technical shift involves the use of vision large language models to interpret user screens directly. This approach eliminates the requirement for every third-party application to write custom integration code to access AI features. Additionally, the new Core AI library facilitates the porting of existing models from Meta’s PyTorch ecosystem to Apple hardware by mapping operations across the FX graph node-by-node. Early adopters on the iOS 27 Developer Beta waitlist are expected to provide the first credible performance data in the coming weeks.
This development matters because it represents a pragmatic evolution of on-device intelligence that addresses the limitations of previous iterations. By leveraging vision capabilities, Apple reduces the friction for software developers while maintaining data privacy through local processing. The integration with established open-source frameworks like PyTorch lowers the barrier for developers to deploy advanced models without rebuilding their entire stacks. Furthermore, the reliance on a licensed Gemini-derived model suggests Apple is prioritising speed and feasibility over building a proprietary foundation model from scratch. If the beta testing confirms these features work as intended, it could mark a turning point where consumer AI becomes genuinely useful rather than a novelty. The success of this strategy will determine whether Apple can regain trust after the overhyped announcements of 2024.
* Apple is using a licensed Gemini-derived model running on private cloud compute rather than a custom-built foundation model.
* Vision LLMs allow Siri to extract information from screens without requiring custom code in every existing app.
* The Core AI library enables developers to port PyTorch models to Apple hardware with minimal friction.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
