Apple's Gemini-powered Siri will run both on-device and in the cloud, relying heavily on Google and Nvidia despite Apple's privacy-focused positioning.

Apple has delayed the AI-enhanced Siri multiple times since first promising it in 2024, and a deal with Google will merge the iconic assistant with Gemini later this year.

The Gemini-infused Siri will run both on-device and in the cloud. On-device AI models are constrained by RAM and processing power—phones can run models with at most a few billion parameters, compared to Google's latest Gemini models, which have trillions of parameters. On-device models are also 'quantized' (reduced in numerical precision) to run faster, which affects accuracy.

This reverses Apple's stated preference for local AI. On Android, Google does not attempt to run Gemini locally as a conversational assistant—it goes straight to the cloud.