On November 7th, according to media reports, Apple is about to reach a cooperation agreement with Google, planning to pay approximately $1 billion (about 7.12 billion yuan) annually to use Google’s AI model, which will be used for a comprehensive upgrade of Siri voice assistant. The new version of Apple’s Siri voice assistant is expected to be released next spring for the iOS 26.4 system.
According to the agreement, Google’s AI model Gemini with up to 1.2 trillion parameters will be used to develop core features such as information summarization and task planning for the new version of Siri. At present, Apple’s cloud based AI model only has 150 billion parameters, with a difference of 8 times in terms of parameters.
While integrating with Google, Apple is also simultaneously advancing its self-developed model and has stated that the self-developed model will be officially launched as early as 2026. However, Gemini 2.5 Pro ranks high on most big model rankings, making it difficult for Apple to catch up with Google.
There are still a few months left until the release, and related plans and collaborations may still change. Apple and Google have no comment on this matter.
1、 Introducing a 1.2 trillion parameter Google model to assist Siri in handling complex tasks
According to foreign media reports, after testing Google’s Gemini, OpenAI’s ChatGPT, and Anthropic’s Claude, Apple has chosen Google to complete the Siri upgrade task, hoping that Google Gemini can serve as a short-term solution before its own model is officially put into use.
This plan to improve Siri through third-party models is internally known as “Glenwood”, led by Mike Rockwell, creator of the Apple Vision Pro headset, and Craig Federighi, head of software engineering at Apple, laying the foundation for a new version of Siri (codenamed “Linwood”) for iOS 26.4, scheduled to be released next spring.
Compared to the currently used 150 billion parameter model, the Gemini model customized by Google for Apple has a parameter count of up to 1.2 trillion, achieving significant improvements in computing power and contextual understanding.
Gemini will be responsible for the summary and planning functions of the new version of Siri, which can help the voice assistant integrate information and decide how to perform complex tasks. Meanwhile, Apple’s internal model will continue to be responsible for some Siri functions.
However, in terms of privacy services, although Apple uses Google’s model, user data will still run on Apple’s own private cloud computing servers to ensure that the data does not access Google. Apple has deployed AI server hardware to support the computation of the Gemini model.
The two companies have maintained a cooperative relationship, and foreign media reports that although the scale of this cooperation is not small, it may not be publicly promoted. In this collaboration, Apple will use Google Vision as the behind the scenes technology supplier, which is different from Apple’s previous plan to integrate Gemini as a chatbot directly into Siri, and also different from Google’s protocol of becoming the default search engine for Safari browser.
During Apple’s earnings conference call on October 30th, CEO Tim Cook stated that in addition to the current ChatGPT option, Siri may offer more chatbot features in the future.
2、 Apple plans to continue developing its own model with 1 trillion parameters, to be used as early as next year
Apple is not the only company that uses Gemini to support AI capabilities. Many large companies are developing based on Google’s Vertex AI platform, such as the American social media company Snap.
Foreign media speculate that this move may indicate that Apple acknowledges that it has fallen behind in the AI field and is now willing to rely on third-party technology to catch up.
Insiders say that Apple does not intend to use Google Gemini as a long-term solution and ultimately hopes to replace Gemini with an internal solution.
In July of this year, Pang Ruoming, former head of Apple’s basic modeling team, was poached by Meta. Despite Apple’s continuous loss of AI talent, the management still plans to continue developing new AI technologies and hopes to eventually replace Gemini with their own solutions.
Apple’s modeling team is developing a model with 1 trillion parameters, hoping to officially apply it as early as next year.
Apple executives believe that the new model being developed can achieve a quality level comparable to the customized version of Gemini. However, Gemini 2.5 Pro ranks high on most big model rankings, and Google is constantly improving Gemini. It is not easy for Apple to catch up.
Apple is also continuing to work hard to introduce Apple Intelligence and the new version of Siri into the Chinese market. Due to the lack of Google services in China, it is unlikely that the upgrade of the Chinese version of Siri will rely on Gemini.
According to foreign media reports, the Chinese version of Apple’s intelligent services may adopt a self-developed model and be paired with Alibaba’s filtering layer. In addition, Apple has been considering collaborating with Baidu to launch AI services in China.
Conclusion: Combining internal and external factors, Apple accelerates the layout of AI
In the fiercely competitive AI market, facing the current situation of internal AI team talent loss and difficulties in developing self-developed models, Apple is leveraging its strategic layout of cooperating with Google to ensure that it does not fall behind in the AI wave.
However, it is still a few months before the release of Apple Siri, and it is unknown how the two tech giants will ultimately collaborate. By relying on external technology to catch up and accelerating independent research and development, the fate of the intelligent voice assistant Siri ultimately lies in Apple’s own hands.
暂无评论内容