Apple Gemini models form the foundation of a new approach the company is taking to develop smaller, on-device artificial intelligence capabilities for Siri, according to a report by The Information published on March 25, 2026.
Google granted Apple complete access to the Gemini model within Google’s own data centers. Apple uses this access for a process called distillation, which involves creating smaller, task-specific models that can run directly on Apple devices without requiring an internet connection.
How Apple Gemini Models Power the Distillation Process
Apple instructs the main Gemini model to perform a series of tasks, generating high-quality results along with a detailed breakdown of the reasoning process. Apple then feeds those answers and reasoning steps into training pipelines for smaller, less resource-intensive models.
The smaller models learn the internal computations used by Gemini as a result of this process. Consequently, they produce outputs with Gemini-like performance while requiring significantly less computing power to operate.
Customization Challenges With Gemini
Apple also holds the ability to edit Gemini directly to ensure the model responds to queries in a manner that aligns with Apple’s requirements. However, the company has encountered difficulties because Gemini was originally tuned for chatbot and coding applications, which does not always match Apple’s specific needs for Siri.
The report notes that Apple’s Foundation Models team continues to develop Apple-native AI models that are separate and distinct from the Gemini models. This parallel development suggests Apple is not fully dependent on Google’s technology for its long-term AI strategy.
Siri’s Planned Capabilities in iOS 27
Apple is relying on Google’s Gemini models for the smarter, chatbot-style version of Siri planned for iOS 27. The updated assistant is expected to handle a broad range of tasks comparable to those supported by other major chatbots.
Specifically, the enhanced Siri will be able to answer questions, summarize information, scan and understand uploaded documents, tell stories, provide emotional support, and complete tasks such as booking travel arrangements.
Context: Apple’s Broader AI Strategy
The partnership between Apple and Google on AI model development represents a notable shift in how Apple approaches on-device intelligence. Apple has historically prioritized building proprietary models, and the Gemini collaboration adds an external layer to that strategy.
Furthermore, the distillation approach allows Apple to maintain user privacy by enabling AI features to function locally on Apple devices, without sending data to external servers. The timeline for iOS 27 and the full rollout of these Siri capabilities has not been confirmed publicly.
“Apple has full access to Gemini to customize the model for Siri and other AI features.”
The Information, March 25, 2026