Apple is planning to use the artificial intelligence model of Google to enhance your voice assistant Crab and is willing to shell out around 1 billion dollars per year to his mother house A, using a customized version of Gemini.
It would be one workaround for Apple, waiting for its models to be sufficiently powerful, but for which it should wait until next spring. In the meantime, the company would have at its disposal from Google a ultra-powerful artificial intelligence model from 1,2 trillion parameters, a measure of the complexity of artificial intelligence software, which would put the level of current Apple models to shame.
Apple reportedly chose Google for Siri's AI functions
Yesterday a Wall Street shares of both companies they briefly jumped to session highs following the news. Apple title gained less than 1%, reaching $271,70, while A gained as much as 3,2%, reaching $286,42.
The Cupertino company had previously evaluated the use of other models as well to address the Siri hardening, reports BloombergBut after testing Gemini, OpenAI's ChatGPT, and Anthropic's Claude, Apple turned to Google earlier this year.
Il Google's customized Gemini system, represents an important step forward compared to the model of Apple Intelligence From the 150 billion parameters currently used for the cloud version, analysts say: the addition of Gemini would significantly expand the system's power and its ability to process complex data and understand its context. Known internally as Glenwood, the project to refine Siri with a third-party model was led by Vision Pro headset creator Mike Rockwell and software engineering manager Craig Federighi. The new voice assistant, expected in iOS 26.4, has code name Linwood.
Apple's team is working on a cloud-based model with 1 trillion parameters that it hopes to deploy in consumer applications as early as next year. Apple executives believe it can achieve a level of quality similar to the custom Gemini offering. But Google continues to improve Gemini, and catching up won't be easy. Gemini version 2.5 Pro tops most rankings comparing large language models, the foundation of generative AI.
