Apple announced today at WWDC 2026 that it will carry out a major architectural upgrade of its Apple Intelligence platform, introducing a basic model developed in in-depth cooperation with Google, and adopting Gemini series of technologies to bring users stronger understanding, reasoning and multi-modal processing capabilities.The new architecture revolves around jointly created Apple Foundation Models, which are specifically adapted to run on local devices or server-side through Apple's existing Private Cloud Compute infrastructure, balancing performance with privacy.
Apple said that this cooperation with Google is a "deep" technical collaboration aimed at bringing "huge upgrades" to Apple Intelligence, including more advanced language understanding and reasoning capabilities, as well as multi-modal capabilities including image understanding and generation. Relying on the upgraded model, Apple Intelligence will support a series of new usage scenarios, such as more realistic image creation, more advanced photo editing, and visual content-based question and answer functions, expanding the application boundaries of the system in the fields of creativity and productivity.
According to reports, on specific devices, users will also receive a higher-order model version with stronger computing power, thereby unlocking additional features, including more natural speech generation, higher-precision dictation transcription, and stronger natural language understanding capabilities. However, Apple has not yet specified in the release which models will support this high-performance version, only emphasizing that it will provide different levels of smart experiences based on differences in device computing power.
At the architectural level, Apple has designed a new system "coordinator" for the new version of Apple Intelligence, which serves as the central component of the entire intelligent function and is responsible for safely coordinating the invocation of different intelligent capabilities on each platform. Apple claims that the coordinator can dynamically schedule and customize the system's response based on the application currently being used and the user's current specific tasks, thereby achieving true "system-level intelligence" and not just local functions within a single application.
At the press conference, Apple also deliberately compared its own path with other manufacturers in the industry, believing that some competitors are "blindly rushing forward" in the artificial intelligence competition without fully considering user rights and privacy protection. Therefore, Apple once again emphasizes that Apple Intelligence will continue its "local first" processing strategy and use Private Cloud Compute to reduce privacy risks as much as possible when cloud computing power is needed to ensure that data is only used to complete the user's current request.
Apple reiterated that under this model, user data will not be accessed or used for other purposes by Apple or third parties, and the cloud processing process is also based on strict technical and institutional constraints. The company also stated that it would allow external experts to independently verify its privacy and security implementation, saying that the relevant privacy commitments "can be tested at any time" in an attempt to respond to external concerns about data security in the era of large models with greater transparency.