Apple is in talks with Google about using Google's Gemini AI as the backbone for the upcoming smarter version of Siri, reports Bloomberg. The discussions are in the early stages, but Apple reportedly approached Google to ask the company to create a custom AI model that would be used for Siri.
Google has started training a model that would run on Apple's Private Cloud Compute servers in response. Apple previously held talks with OpenAI and Anthropic, and both of those companies have been working on versions of their models for Apple's servers.
While Apple is exploring partnerships with different AI companies to power an improved version of Siri, Apple still has not made a decision on whether it will use a third-party AI solution or go with the LLM models it has been developing in-house.
Apple is testing multiple LLMs, including its own, to determine which will provide the best results to customers. There are two versions of the new Siri in development, including one that is powered by Apple's own models and one that runs on third-party models.
Apple has been developing a more advanced version of Siri that overhauls the entire Siri architecture. Apple's software engineering chief Craig Federighi called this the "second-generation" Siri architecture earlier this month, and he said that the "end-to-end revamp" of Siri brought the results that Apple needed for more advanced AI features, such as the personalization that it promised as part of Apple Intelligence in iOS 18.
"This has put us in a position to not just deliver what we announced, but to deliver a much bigger upgrade than that we envisioned," Federighi said.
Apple was leaning toward partnering with Anthropic, but Anthropic's fees prompted Apple to consider other AI companies as well. No deal has been inked with Google, OpenAI, or Anthropic as of yet. Rumors suggest that LLM Siri will launch in spring 2026.
