Chrome AI features reported to use 4GB RAM

Google Chrome does download on-device AI models in the background for some browser and web features, according to Google’s own support and developer documentation. The recent claim that Chrome “uses 4GB RAM for AI” appears to mix together several different things: model download size, storage use on disk, hardware eligibility, and live memory use while an AI feature is running. Google’s public documentation confirms background downloads and references a larger Gemini Nano variant such as 4B, but it does not say Chrome universally consumes 4GB of RAM at all times. That leaves the headline figure as a reported user observation rather than a blanket specification from Google. ### Is the 4GB figure about RAM, storage, or the model itself? Google’s October 2025 developer note says Chrome can choose between “a larger, more capable Gemini Nano variant (such as 4B parameters)” and a smaller 2B variant, depending on the device. That language refers to model size and selection, not a promise that 4GB of active memory will always be used. The same page says the model is downloaded on demand and managed automatically in the background. (developer.chrome.com) Google’s “Get started with built-in AI” page also separates storage from runtime requirements. It says users need at least 22 GB of free space on the volume containing the Chrome profile, while CPU-based use requires 16 GB of RAM or more and at least four CPU cores. Those are eligibility thresholds for supported devices, not a statement that Chrome will permanently reserve 4GB of memory. (developer.chrome.com) ### When does Chrome actually download these AI models? Google Help says Chrome “may download on-device Generative AI models in the background” so features that rely on them stay ready for use. The same help page says those models can support tasks such as writing help, scam warnings, page summaries and tab organization. Users can disable that behavior by turning “On-device AI” off in Chrome’s System settings. (developer.chrome.com) Google’s developer documentation gives a narrower technical trigger for built-in AI APIs. It says the initial Gemini Nano download is triggered by the first call to a `*.create` function for an API that depends on Gemini Nano, and in some cases an `availability` call can also trigger a download shortly after a fresh profile starts if scam detection is active. That means the model is not necessarily fetched for every user immediately on install, but it can arrive without a separate prompt once the relevant feature path is hit. (support.google.com) ### Which Chrome AI features are tied to these models? Google’s support pages list several AI features that can rely on on-device models, including help with writing, page summarization, scam detection and tab organization. A separate help page for Gemini in Chrome says the feature can summarize articles, explain topics, compare information across pages and share up to 10 open tabs with Gemini on desktop computers. (developer.chrome.com) Google also says not every AI feature in Chrome depends on an on-device generative model. Its support page notes that some AI features may still work even if on-device generative models are removed, which is another reason the 4GB claim should not be treated as a single number covering every Chrome AI tool. ### So what can users verify on their own machines? Google’s support documentation says users can manage these downloads in Chrome by going to Settings, then System, and turning “On-device AI” on or off. (support.google.com) The same page says deleting the models frees storage space, but features that rely on them will stop working until the setting is turned back on and the files are downloaded again. Google’s developer page says installed model versions can be checked manually at `chrome://on-device-internals`. For users trying to sort out whether a reported spike is disk use, download size or live memory consumption during an AI task, that internal page and the System setting are the clearest first checks Google currently documents. (developer.chrome.com) (support.google.com)