Apple is working on a comprehensive redesign of its voice assistant Siri, and is preparing to use Google's artificial intelligence capabilities via the Gemini model to speed up the development process, in a deal that could cost the company around $1 billion a year.

According to a new report from Bloomberg, Apple is close to signing an agreement with Google to use the Gemini model within its AI feature package Apple Intelligence, specifically in the personal assistant Siri with its new version officially scheduled to launch in 2026 after several postponements.

The sources indicate that Apple intends to rely on one of Google's Gemini models, which includes about 1.2 trillion parameters, which is far more than Apple's current models, which are only close to 150 billion parameters. This huge difference gives Siri the ability to process more complex data and perform more sophisticated tasks. Parameters are a unit of measurement for the performance and capabilities of AI models.

Even with the large size of the model, Siri's role will be relatively limited. According to the report, Gemini technologies will be used only for summarizing and planning tasks, which require understanding the context and dealing with multiple tasks. The rest of Siri's functionality will still rely on Apple's internal models for privacy.

Apple: Privacy First

Gemini technologies are set to work on its own servers, which it calls Private Cloud Compute, to ensure that users' data remains completely isolated from Google's infrastructure. As such, Google's role will remain limited to a cloud partnership without direct integration into the iPhone's operating system.

Gemini is not expected to integrate directly into Siri's chat system at this time, although CEO Tim Cook has previously hinted at the possibility of expanding Siri's support for third-party models alongside ChatGPT in the future.