Apple is preparing to unveil the most thorough rebuild of Siri since the assistant launched in 2011 at its Worldwide Developers Conference on June 8, three people familiar with the plans said this week. The new assistant — internally codenamed "Tempo" — will be powered by a Gemini-based model running on Google Cloud, under a multi-year partnership that Apple and Google formalised in January and that Google confirmed publicly last month.
The redesign brings Siri visually and functionally closer to a modern chatbot, with a rebuilt conversational interface, deeper system integration that allows it to act on app data across the OS, a dedicated standalone app, and an always-available query input from anywhere on the iPhone. Developers will be able to integrate their own AI models with Siri, in a structure analogous to the OpenAI integration Apple shipped last year.
The timeline puts an internal Apple build into a final pre-WWDC freeze this week, with public beta available in early July and the wider iOS 27 rollout in September. Macintosh, iPad and visionOS releases follow the same cadence.
The shift to Gemini is a significant strategic concession for Apple, which spent four years pursuing an entirely in-house Apple Intelligence stack and publicly committed to that path under former AI chief John Giannandrea. Senior Apple executives told staff in March that the company's own foundation models will continue to handle on-device tasks but that Gemini's 2026 generation outperformed Apple's server-side equivalents by margins large enough to make the partnership "the responsible call."
Apple has not officially confirmed the architecture details and is expected to position the partnership as one of several models a user can choose, alongside OpenAI's ChatGPT and an in-house Apple Intelligence option for privacy-sensitive tasks. The final reveal is being held tightly: even the keynote's opening minutes are reportedly being storyboarded by Tim Cook personally.