Apple has reportedly abandoned its in-house development efforts for the core Siri model and has instead contracted Google to design a bespoke version of its Gemini Large Language Model (LLM) specifically for deployment within Apple’s Private Cloud Compute framework, making Apple’s AI more on par with its competitors.
According to a recent dispatch in the pay-walled Power On newsletter by Bloomberg’s Mark Gurman, Apple has ended plans to develop its own Siri model, a component of the broader Apple Intelligence features first promoted in 2024, according to the Bangkok Post.
The technology firm will now pivot to utilising the customised Gemini model engineered by Google for more complex tasks.
The processing architecture for the upgraded Siri is understood to be bifurcated into two primary operational tiers:
General Commands: For routine queries, Siri will process requests on-device utilising either Apple’s proprietary AI model or a third-party model like ChatGPT.
Complex Commands: More intricate requests, such as those requiring cross-application coordination, will be routed for server-side processing via Apple’s Private Cloud Compute, employing the Google-designed Gemini model.
Apple maintains a stringent focus on user security and privacy, ensuring the upgraded Siri maintains its existing high standards. All data transmitted to the cloud for processing will reportedly be encrypted and handled using a stateless approach, ensuring no residual user traces remain on the servers.
Mr Gurman said that Apple is planning an official unveiling of this new Siri system around March 2026, likely coinciding with the introduction of new smart home devices.
However, the rollout for the Chinese market is facing an indefinite delay due to hurdles including navigating complex legal requirements and securing necessary governmental approvals within the jurisdiction.
