Today, Apple announced significant architectural upgrades to its Apple Intelligence platform. The revamped system is built on foundation models co-developed with Google, leveraging core technologies behind the Gemini series. According to Apple, the new architecture centers on jointly developed Apple Foundation Models, which can operate both locally on devices and server-side via Apple’s existing Private Cloud Compute infrastructure. Describing this collaboration as "deep," Apple stated it brings a "massive upgrade" to Apple Intelligence, significantly enhancing comprehension and reasoning capabilities while adding multimodal support such as image understanding and generation. The upgraded models unlock several new features, including realistic image creation, advanced photo editing, and visual question-answering abilities. Some devices will receive higher-performance versions supporting voice generation, more precise dictation, and stronger natural language understanding—though Apple did not specify which devices qualify. At the heart of the new architecture lies a System Orchestrator responsible for securely coordinating various AI functions across Apple’s entire ecosystem. Apple noted that this orchestrator customizes responses based on currently running apps and specific user tasks, enabling true whole-system intelligence. In its announcement, Apple deliberately distanced itself from competitors, criticizing some companies for rushing ahead without regard for users. It reiterated that Apple Intelligence relies on on-device processing and Private Cloud Compute, ensuring user data is used solely for executing current requests and remains inaccessible to Apple or any third parties. Additionally, Apple emphasized that external experts could verify these privacy commitments at any time.