No, Google hasn’t changed Chrome’s local AI features—it’s just as confusing as ever

Google continues to enhance its products with artificial intelligence, and Chrome is no exception. The browser has introduced several AI functionalities, including split-screen support for the Gemini chatbot and automated web browsing capabilities. However, some users have recently observed that Chrome seems to require more storage for these features. Indeed, Chrome has been downloading a 4GB AI model for on-device processing, a practice that has been in place for years. Despite this, Google has not made any significant changes to the AI features used locally in Chrome. The confusion arises from the company's lack of clear communication regarding its AI developments. Recently, a user noticed the sudden appearance of the 4GB Gemini Nano model and speculated that Google was rolling out this AI across all Chrome installations. However, this inference is somewhat misleading. In 2024, Google announced plans to integrate local AI capabilities into Chrome, which would enhance functionalities such as Help Me Write, tab organization, and scam detection. Although the potential applications have evolved, Google has not implemented these models universally. It's possible that users’ machines have been running the Gemini Nano model since its initial release in 2024, or it may have been downloaded recently. Google clarified to Ars Technica that the criteria determining whether the Gemini Nano model is active on a device are complex, involving hardware specifications, account features, and previous visits to sites utilizing Google’s on-device Gemini API. As Gemini Nano frequently appears for the first time on various machines, it can create the impression that this feature is newly introduced. However, Google has confirmed that the 4GB size of Chrome's Gemini Nano model has remained consistent since its launch two years ago.