Apple Announces Google Gemini-Powered Siri Overhaul With New Two-Tier On-Device Models and Stricter Hardware Requirements

Apple introduced its long-delayed Apple Intelligence update for Siri during its pre-filmed Worldwide Developers Conference keynote. The new Siri AI will roll out with operating system updates this fall and will include tighter integration across Apple’s platforms. ” Scripted demos showed Siri handling multi-step tasks.

In one sequence a user asked about World Cup schedules, requested recipes for a Brazil versus Morocco match, located a dessert mentioned by friend Maria in Messages, compiled a watch-party menu, and sent an invitation to a group chat. In another, Siri identified a photo location, found friend Jeff’s new address, and routed directions to the arch with a stop at Jeff’s via Apple Maps.

Siri AI can read on-screen content, search across messages and email, and customize email drafts to match prior writing style for the same recipient.

It can also pull world knowledge through private cloud compute, trigger app actions, and extract multiple calendar reminders from a single concert-schedule image. Visual intelligence lets users ask questions from the camera app or, on visionOS, about physical objects in view.

A new “Write with Siri” feature generates text anywhere typing is supported, and an automatic proofreading system checks style system-wide.

On macOS the features integrate with Spotlight search and allow ctrl-click queries on photos, files, or multiple items. On visionOS a glowing animated Siri Orb can be placed in the workspace. Conversations will be stored locally and in iCloud inside a dedicated Siri app that lets users continue a session across devices.

Earlier this year Apple signed a multi-year agreement to base the update on Google’s Gemini model. Apple also announced a two-tier on-device model structure. The most capable model requires iPhone Air or iPhone 17 Pro, iPads with M4 or later chips and at least 12 GB of memory, or Macs with M3 or later chips and at least 12 GB of memory.

Devices that meet current Apple Intelligence requirements but lack these specifications will receive a less capable model without the new expressive Siri voice or the dictation accuracy improvements. Many features will continue to use private cloud compute servers. Image-generation tools will carry daily usage limits, with expanded allowances for most iCloud+ subscribers.

The Siri AI features will launch in English only and later expand to every language supported by iOS. Apple stated that the features include “non-negotiable” privacy protections ensuring Siri AI conversations remain unavailable to Apple or any other party.