Apple explores Google partnership to power next-generation Siri

Apple is exploring partnerships with Google to power the next-generation Siri, with the tech giant conducting an internal “bake-off” between different AI models to determine the best approach. This development comes as Apple faces significant delays in delivering its promised AI-powered Siri redesign, originally announced for iOS 18 but now pushed to spring 2026.

What you should know: Apple has approached Google to explore building a custom Gemini-based AI model that would run on Apple’s servers to power the new Siri experience.

• Google has already begun training a model specifically designed to operate on Apple’s infrastructure, according to Bloomberg’s Mark Gurman.
• The discussions remain private and no final decision has been made about which AI provider will ultimately power Siri.

The big picture: Apple is simultaneously developing two versions of the new Siri to compare performance and capabilities.

• One version, codenamed “Linwood,” is powered entirely by Apple’s own AI models.
• Another version, dubbed “Glenwood,” runs on external AI technology from potential partners.
• This dual-track approach allows Apple to evaluate the best solution before committing to a specific direction.

Who else is involved: Multiple AI companies are competing for the opportunity to power Siri’s next iteration.

• Anthropic, maker of the Claude AI model, has reportedly been viewed as Apple’s preferred choice but may be asking too high a price.
• OpenAI, the company behind ChatGPT, was also previously reported as a potential partner under consideration.
• Google’s Gemini already powers Android’s AI system and Samsung’s AI features across various platforms.

Key technical developments: Apple is making significant investments in its own AI capabilities while exploring partnerships.

• The company has begun testing its first trillion-parameter model, which would be substantially more powerful than its current 150 billion-parameter models.
• Some Apple leaders have considered using non-Apple models for broader Apple Intelligence features, though the company is currently only actively seeking a Siri model provider.

In plain English: Parameter count is essentially a measure of an AI model’s complexity and capability—think of it like the difference between a pocket calculator and a supercomputer. Apple’s current models use 150 billion parameters, while their new experimental model uses one trillion parameters, making it roughly six times more sophisticated and potentially much better at understanding and responding to complex requests.

Why the delays matter: Apple’s Siri overhaul has faced multiple setbacks that have pushed back its original timeline.

• The company first announced an all-new Siri experience in June 2024, promising delivery with iOS 18.
• Major roadblocks, leadership reorganization, and the decision to start over with a new system have delayed the project significantly.
• Apple is now targeting a spring 2026 release for the redesigned Siri experience.