Apple Unveils New AI Architecture Built on Google Gemini Models

A major overhaul of Apple's Intelligence platform has been announced, with a new architecture centered around foundation models developed in collaboration with Google. The technology behind the Gemini family is at the heart of these co-developed models, which can run both on-device and on servers through Apple's Private Cloud Compute infrastructure.

The upgraded models bring significant capabilities to Apple's platforms, including state-of-the-art understanding and reasoning abilities as well as multimodal support for image understanding and generation. This means users will be able to create realistic images, edit photos with advanced tools, and engage in visual question answering.

Devices set to receive a higher-power version of the model will gain additional capabilities such as speech generation, improved dictation accuracy, and stronger natural language understanding. However, Apple has not specified which devices qualify for this upgrade.

A new system orchestrator sits at the center of the revised architecture, coordinating features across Apple's platforms securely. This allows the system to tailor its responses based on the active app and user task, enabling truly system-wide intelligence.

Apple framed its approach as a contrast to competitors that 'racing forward' without regard for users. The company reiterated its commitment to on-device processing and Private Cloud Compute, ensuring user data is only used to execute immediate requests and remains inaccessible to Apple or third parties. Outside experts can verify these privacy guarantees at any time.

The new architecture marks a significant upgrade for Apple's Intelligence platform, bringing advanced capabilities that were previously unavailable.