原文 · 未翻译
Tech
AI
News
Chrome’s AI features may be hogging 4GB of your computer storage
Here’s how you can find out, and get that storage back if you need it.
Here’s how you can find out, and get that storage back if you need it.
Google Chrome may be taking up more of your storage than expected thanks to a large on-device AI model file that, in some cases, is being automatically downloaded to the browser’s system folders. Users who have noticed unexplained drops in their available desktop device storage are now discovering that Chrome is installing a 4GB weights.bin file inside their browser directory when certain AI features are enabled.
The weights.bin file in question is connected to Google’s Gemini Nano AI model, which powers Chrome AI tools like scam detection, writing assistance, autofill, and suggestion features. As the Gemini Nano model is designed to run locally, it needs to use training parameters stored on your device rather than pulling information from cloud-based models. That provides some privacy benefits, but isn’t ideal if you’re low on storage — especially as users aren’t being clearly notified about the file size requirements.
If you have certain Gemini AI features enabled on Chrome, it’s likely that the 4GB file has already been downloaded to your system. You can check by opening your Chrome data folders and inspecting the OptGuideOnDeviceModel directory for the weights.bin file.
You can’t simply delete this to free up space, however — if you still have AI features enabled, Chrome may re-download it again in the future. That means you’ll need to head to Settings>System and toggle off the On-Device AI option to remove those features and prevent the file from coming back.
Google does specify that “Gemini Nano’s exact size may vary as the browser updates the model,” but this information is presented in a lengthy guide for built-in AI features rather than at the point of enabling them in Chrome. If Google had made the storage requirements clearer to users — or provided an option to power Chrome AI features with cloud-based models — this confusion could have been avoided.