Apple will use Google’s Gemini models to power new Siri release – News

Apple Inc. will use Google LLC’s Gemini series of large language models to power its Siri voice assistant.

The companies announced the partnership, which is described as a multi-year deal, this morning. When rumors of the collaboration first emerged in August, Bloomberg reported that Apple was set to pay Google about $1 billion per year to access Gemini. Sources told the publication that the iPhone maker sought a custom Gemini model with 1.2 trillion parameters.

According to today’s announcement, Apple will deploy not one but multiple Gemini models. It will use them to power a new version of Siri set to roll out later this year. According to Apple, the upgraded assistant provides more personalized responses and can perform multi-step tasks spanning multiple apps.

The company’s decision to use multiple Gemini models rather than just one might be motivated by economic reasons. Some applications use a large, processor-intensive model to answer users’ most complicated prompts and relegate simpler requests to a lighter algorithm that uses less hardware. That arrangement helps drive down infrastructure costs.

Gemini 3, the latest iteration of Google’s LLM series, is available in multiple versions with varying performance and price points. The most capable model, Gemini 3 Pro, costs four times more than the entry-level Flash edition.

Apple stated that it will use the LLM series alongside Google’s “cloud technology”. That hints the iPhone maker may use Google Cloud to power parts of its Gemini deployment.

The upgraded version of Siri that will use the LLM series is part of an artificial intelligence feature suite called Apple Intelligence. In 2024, Apple partnered with OpenAI Group PBC to integrate ChatGPT into the suite. Currently, Siri uses the chatbot to answer some user prompts. An Apple spokesperson told CNBC today that it won’t change its partnership agreement with OpenAI. 

The iPhone maker stated that Apple Intelligence will “continue to run on Apple devices and Private Cloud Compute.” Private Cloud Compute is a set of cybersecurity features the company debuted in 2024. It deletes the user data that Siri uses to answer prompts after it’s no longer needed. Furthermore, the servers that run Private Cloud Compute lack many of the troubleshooting tools typically used by data center operators. Apple says that doing away with those components reduces the risk of cyberattacks.

Today’s partnership announcement pushed the market capitalization of Google parent Alphabet Inc. above $4 trillion, making it the fourth company to have achieved the milestone. It was preceded by Apple, Nvidia Corp. and Microsoft Corp. last year.

Image: Google