With the official unveiling of iOS 27 at the Worldwide Developers Conference (WWDC 26), Apple's deeply integrated artificial intelligence system - Apple Foundation Models (AFM) - has become the focus of the industry. In response to external doubts about "rebranding Google Gemini," several Apple executives gave clear and firm responses in recent media interviews: Apple Foundation Models are not a simple rebranding of Gemini, but rather a completely self-controlled achievement by Apple.
Apple clearly stated that although some Gemini technology was indeed referenced during the training and distillation of the model, the final product delivered to users features core architecture, underlying code logic, technical implementation paths, and supporting data systems entirely built by the Apple team. This statement aims to dispel user and developer concerns, emphasizing Apple's core control and ecological independence in the AI field.
Currently, Apple's AFM system has shown a clear hierarchical structure to handle tasks of varying complexity:
On-device models, AFM Core focuses on handling basic AI local operations; the more advanced AFM Core Advanced introduces a sparse architecture, capable of supporting native multimodal capabilities, carrying more complex computing needs on local devices.
Cloud-based models are divided according to workload requirements: AFM Cloud handles high-load requests that are difficult for devices to process; AFM Cloud Image specializes in high-precision image generation and editing; while AFM Cloud Pro, as the most advanced, is dedicated to agent tools and the heaviest computing tasks.
To ensure that every device with Apple Silicon can perform at its best, these models are deeply customized for the chip. Apple emphasized that when users access related services, the system architecture ensures that users do not directly access Google's code, Gemini agents, or Google Search, thereby ensuring the purity of user experience and privacy security to the greatest extent possible.
Even in the deployment of cloud infrastructure, Apple maintains a very high security standard. Although the infrastructure for AFM Cloud Pro uses Google Cloud servers and NVIDIA GPU resources, the entire processing workflow still strictly follows Apple's "Private Cloud Compute" certification system. This means that even if the task is completed in the cloud, Apple ensures strict control over privacy protection.
As iOS 27 is gradually rolled out, this AI matrix driven by Apple's self-developed underlying technology is trying to find a new balance between "edge-cloud collaboration" and "privacy compliance," demonstrating Apple's unique path in building an AI ecosystem to global users.
