Apple to Use Google Gemini to Fix Siri

Mark Gurman reported today that Apple is close to finalizing a deal with Google to use its Gemini AI models for Siri.

Under the terms of the deal, Apple will pay $1 billion per year to use a high-end Gemini AI model with 1.2 trillion parameters to help bring its Siri assistant into the Apple Intelligence era. The shift comes after a months-long testing period and is separate from the failed 2024 tests that would have seen Gemini come to the iPhone and other Apple devices more broadly.

Apple has infamously stumbled in bringing Apple Intelligence features to Siri after announcing them at WWDC 2024. Since that announcement, the company has built out various AI-capabilities under the Apple Intelligence umbrella, but Siri remains as pointless as ever, with Apple continually delaying the conversational rollout.

Apple reportedly also tested OpenAI ChatGPT and Anthropic Claude for Siri, but it found Gemini to be better for its needs. Apple being Apple, it will of course replace Gemini with its own models if and when they’re sophisticated enough. But the plans and partnership could still change in the interim, Gurman notes, and with Google’s $20+ billion payoff to Apple for Search placement still intact, all bets are off: $1 billion per year is pocket change by comparison.

Gurman says that Apple plans to use Gemini in Siri for summarizing and planning functionality, while some other Siri features will still use Apple’s models. Gemini for Siri will run on Apple’s Private Cloud Compute infrastructure, and Apple has already allocated the hardware needed to power the model.

Apple now plans to ship conversational Siri in the first half of 2026.