June 11, 2026
At WWDC 2026 on June 8, Apple unveiled the next generation of Apple Intelligence and a rebuilt Siri, now called Siri AI, that runs on a custom 1.2-trillion-parameter Google Gemini model hosted inside Apple’s Private Cloud Compute. Apple is paying Google a reported 1 billion dollars per year for the custom Gemini model, which handles Siri’s most demanding reasoning while Apple’s own models stay on the device. The release is Apple’s answer to years of criticism that Siri had fallen behind ChatGPT, Gemini, and Claude.
What can the new Siri AI do?
Siri AI becomes a standalone app with a single Search or Ask field, similar to ChatGPT, Gemini, and Claude, and supports both text and voice conversations. On iPhones with a Dynamic Island, a Siri animation appears there while a request is being handled. The assistant gains personal context, meaning it can act on a user’s emails, photos, files, and messages, along with on-screen awareness and deeper cross-app actions.
Apple is moving developers from the older SiriKit framework to App Intents so that apps can expose actions Siri can chain together. In Apple’s demonstration, a single spoken request could find a specific photo, edit it, and attach it to a message without the user switching apps. This positions Siri as a persistent interface for getting tasks done rather than a one-off question and answer assistant.
How the Gemini model and Private Cloud Compute work together
Siri AI uses a three-tier routing system. Simple requests stay on the device and run on Apple’s own updated foundation models. Moderately complex requests go to Apple’s Private Cloud Compute servers. The heaviest reasoning tasks route to the custom 1.2-trillion-parameter Gemini model running on Google Cloud, on Nvidia Blackwell B200 GPUs.
Apple has stressed that its own on-device foundation models contain none of Google’s Gemini code, and that Gemini is used only for the cloud reasoning layer. Apple routes that traffic through its Private Cloud Compute infrastructure, which is designed so that personal data is not retained and is not visible to Google. The underlying deal, first reported in January 2026, has Apple paying Google roughly 1 billion dollars per year for the custom model.
Choosing Claude, ChatGPT, Gemini, or Grok inside Siri
Apple is expanding beyond its earlier ChatGPT handoff with a new Extensions system in iOS 27, iPadOS 27, and macOS 27. Through Settings, users can set Claude, ChatGPT, Google Gemini, or Grok as their preferred model across Apple Intelligence features, not just for one-off questions.
This turns Apple Intelligence into a marketplace layer where the on-device experience stays consistent while the underlying model is user-selectable. For organisations, it means a single Apple Intelligence workflow can be pointed at the model that best fits a given capability, cost, or compliance requirement, with Anthropic’s Claude, OpenAI’s ChatGPT, Google’s Gemini, and xAI’s Grok all available as options.
How does Siri AI compare to the old Apple Intelligence, Alexa+, and Gemini?
The 2024 version of Apple Intelligence promised a more personal Siri that Apple repeatedly delayed, and the assistant lagged behind ChatGPT, Gemini, and Claude on reasoning and multi-step tasks. By licensing a 1.2-trillion-parameter Gemini model, Apple is closing that gap with the same class of frontier model that powers Google’s own assistant, rather than waiting for its in-house models to catch up.
The move puts Siri AI in direct competition with Amazon’s Alexa+ and with Gemini on Android and Pixel devices. Apple’s main differentiator remains on-device processing for simple tasks and the Private Cloud Compute privacy model for everything routed off the device, which Apple is using to argue that a Gemini-powered Siri does not mean handing user data to Google.
Apple Intelligence and Siri AI availability and pricing
Developer betas of iOS 27, iPadOS 27, and macOS 27 Golden Gate are available after the keynote, with public betas in July 2026 and a general release in the autumn of 2026. Siri AI and the next generation of Apple Intelligence are free features of those operating systems, with the Gemini cloud costs covered by Apple’s reported 1 billion dollar per year agreement with Google.
Alongside the AI news, Apple said apps launch up to 30 percent faster, photo previews load up to 70 percent faster, and file transfers on iPadOS are up to five times quicker. The keynote was also Tim Cook’s farewell as chief executive and included a first preview of homeOS.
Full details are in Apple’s WWDC 2026 announcement.
