Apple Restructures Siri with Trillion-Parameter Gemini Model and Nvidia Confidential Computing

According to Beating monitoring, Apple is restructuring Siri at the foundational level and plans to introduce Google's trillion-parameter Gemini model in iOS 27. The new architecture employs hybrid edge-cloud processing: Gemini undergoes on-device distillation, allowing iPhones to handle basic tasks locally and reduce latency, while complex inference and generation tasks are routed to cloud servers. Apple is incorporating Nvidia's Confidential Computing platform to ensure all cloud-processed data remains encrypted, balancing computational demands with user privacy protection.

The updated Siri will offer greater openness, supporting third-party AI agent integration and enabling bidirectional data flow. Apple is also optimizing system-level resource orchestration to mitigate the impact of large language models on battery life.