Apple is reportedly close to finalizing a major licensing deal with Google that would bring the Gemini language model to Siri, according to Bloomberg. The agreement, valued at about $1 billion per year, would give Apple access to Gemini's capabilities to power core functions of its voice assistant.
At the center of this plan are new "summarizer and planner" features - the ability to efficiently collect, condense, and organize complex information. The Gemini model Apple intends to use reportedly has around 1.2 trillion parameters, making it far more powerful than the 150-billion-parameter cloud AI Apple currently employs. The move follows internal setbacks that caused Apple to miss the initial wave of generative AI adoption.
The new Siri, internally known as "Glenwood" and marketed under the codename "Linwood," is scheduled to debut with iOS 26.4 in spring 2026. The project is led by Mike Rockwell, who also heads Vision Pro development, and Craig Federighi, Apple's software chief. Some Siri components will continue to rely on Apple's own AI technology.
Even though Google is providing the model, it will run on Apple's servers. According to Bloomberg, all user data will remain entirely separate from Google's infrastructure. The integration is designed to happen quietly in the background, and Apple does not plan to publicly market the Google name - unlike the long-standing deal that makes Google the default search engine in Safari.
Apple is still building toward its own models
Before settling on Google, Apple reportedly explored partnerships with OpenAI (ChatGPT) and Anthropic (Claude) but ultimately chose Gemini. Earlier talks about embedding Gemini as a full chatbot failed, and Google's search service is not part of the current agreement.
In the long run, Apple intends to rely on its own AI models again. A 1-trillion-parameter model is already in development and could be ready for consumer devices as soon as next year. The company hopes the system will eventually reach Gemini-level quality. However, Apple has lost several key AI researchers to Meta in recent months, which could slow progress.
In China, Apple will reportedly take a different approach. Gemini is unlikely to be used there, instead, Apple plans to deploy its in-house models filtered through an additional Alibaba mechanism required by Chinese regulators. The company is also considering a separate partnership with local AI firm Baidu.
