Apple unveils rebuilt Siri powered by Google Gemini at WWDC
The redesigned assistant pairs on-device processing with a custom, roughly 1.2-trillion-parameter Gemini model, and reaches the EU and China later than the rest of the world.
- Models & capabilities
- Labs & people
- Notable
Apple opened its Worldwide Developers Conference by unveiling the rebuilt Siri it had signalled five months earlier, when it selected Google’s Gemini to power its next-generation assistant under a reported roughly $1 billion-a-year deal. The new Siri runs a custom Gemini model — reported at around 1.2 trillion parameters — for cloud reasoning, while routing privacy-sensitive requests through on-device processing or Apple’s Private Cloud Compute infrastructure, which Apple said keeps Google without access to user data.
The redesign gives Siri a standalone app, a system-wide “Search or Ask” gesture, on-screen awareness of what a user is looking at, cross-app actions, and access to personal context across messages, email and photos — features closer to a general-purpose assistant than the command-based Siri Apple had shipped for over a decade. Developer betas began immediately, with a public beta due in July and general release alongside iOS 27 expected in the autumn; Apple said the EU launch would be delayed over Digital Markets Act compliance and that China would wait on separate regulatory approval.
The reveal confirmed the strategic reversal signalled in January: rather than compete on a proprietary foundation model, Apple chose to treat frontier AI capability as licensable infrastructure and compete instead on integration, privacy architecture and distribution across its device base — a wager that hundreds of millions of iPhones make Apple, not Google, the platform of record for the assistant even though Google supplies the model underneath it.