Google has quietly updated its support documentation to clarify just how much storage Chrome needs to install on-device AI models. The company now says that downloading the machine learning models that power Chrome's local AI features requires approximately 20 gigabytes of free disk space. That is a significant amount for anyone running a laptop or desktop with a modest SSD, especially as modern operating systems and applications already consume a large portion of available storage.
The update was likely driven by the confusion and frustration some users experienced earlier this year when Chrome began automatically downloading a large file called weights.bin to support Gemini Nano features running directly in the browser. Many people noticed a sudden multi-gigabyte download appearing in their system file managers, often without any clear explanation. Google's support page now sets expectations more explicitly, although the company still uses somewhat vague language about hardware requirements.
What exactly are Chrome's on-device AI features?
Chrome's on-device AI features are part of Google's broader push to integrate Gemini, its AI assistant technology, into the browser. Gemini Nano is a lightweight version of the company's large language model, designed to run on consumer hardware rather than in the cloud. In Chrome, this local model powers a range of experimental tools, including a smart writer that helps compose text, a tab organizer that groups similar tabs, and native text summarization for articles and PDFs. These features are intended to work offline, sending no input data to Google's servers, which appeals to users who are concerned about privacy or have unreliable internet connections.
Running AI locally has a clear privacy advantage: everything stays on the device. But it comes with a hidden cost in terms of system resources. Large language models, even compact ones like Gemini Nano, require substantial storage and memory. The model needs enough space not only for its final installation but also for temporary files, compression artifacts, and potential updates. That is likely why Google requires about 20GB of free space even though the installed model is only around 4GB. The download process may need extra room for verification, staging, and fallback copies, and Google probably wants to avoid failures that could corrupt the browser or leave partial files behind.
Why 20GB feels like a lot
For users with large SSDs, 20GB might be a minor concern. But many laptops, especially Chromebooks and budget Windows machines, ship with 128GB or 256GB of storage. A 20GB requirement represents roughly 8 to 16 percent of the entire drive. On such systems, the operating system, system applications, and user files can fill the disk quickly. Users with only a few gigabytes of free space are already used to warning messages and slowdowns. Adding another 20GB requirement means Chrome may refuse to enable on-device AI operations unless the drive is nearly empty.
It is also worth putting the number in perspective. A standard Chrome installation, without on-device AI models, often occupies around 1.5GB of disk space. The on-device AI models thus increase Chrome's storage footprint by roughly four to five times once installed, and the temporary download requirement is more than thirteen times the browser's baseline. That is a substantial resource commitment for a single application, especially when many users run multiple browsers, communication apps, and creative tools simultaneously.
Google's requirements and automatic download behavior
The support documentation states that Chrome will only attempt to download on-device AI models if the device meets several conditions. In addition to the 20GB free space requirement, the browser needs an unmetered data connection. This is particularly important for users on mobile hotspots or capped broadband plans, as a multi-gigabyte download can consume a large portion of a monthly data allowance. The documentation also mentions that the device must have sufficiently strong performance. Google does not yet provide specific CPU, memory, or GPU thresholds, leaving many users to wonder whether their hardware qualifies.
When Chrome does decide the hardware is suitable, it automatically starts downloading the model files in the background. This automatic behavior surprised many users when it first rolled out. Some found the process intrusive and difficult to stop, especially since simply deleting the weights.bin file did not prevent Chrome from downloading it again. Google has since clarified that the correct way to disable these AI features is to turn off the on-device AI toggle in the browser's settings.
To turn off the features, users need to open Chrome's settings, navigate to the System section, and find the toggle labeled On-device AI. Disabling this toggle will remove already downloaded local models and prevent Chrome from re-installing them. Google warns that manually uninstalling the model files without using this setting is ineffective, as the browser will simply download them again at the next opportunity. For enterprise administrators, Chrome also offers group policies that can control AI features across managed devices, which can prevent unwanted bandwidth and storage consumption in corporate environments.
Implications for privacy-conscious users
While the storage requirement may be annoying, the on-device AI features do offer meaningful privacy benefits. Because the models run locally, text that is analyzed for summarization or composition never leaves the device. This stands in contrast to cloud-based AI tools, which typically transmit user content to remote servers for processing. For business users handling confidential documents, local processing reduces the risk of data leakage and helps meet compliance requirements in industries with strict data protection rules.
However, these benefits are not necessarily obvious to the average user. Many people are unlikely to volunteer 20GB of storage and potentially hours of bandwidth for a feature they may not have chosen to enable. Google has attempted to address this by making the on-device AI toggle available in settings, but the initial auto-download behavior left many users with a sour impression. Clearer communication about the storage requirement and a more explicit opt-in mechanism might have reduced the backlash.
Another consideration is that Chrome's on-device AI features may continue to grow in both scope and size. Gemini Nano is still an early-stage model, and Google is actively expanding the types of tasks it can perform. Future updates could introduce larger model variants or additional components, potentially pushing the storage requirements even higher. Users who are short on space now may find the issue worsening over time, particularly if the default download behavior continues.
The decision to require 20GB of free space undermines a key selling point of lightweight laptops and Chromebooks, which are often marketed as low-cost, low-friction alternatives to traditional computers. These devices frequently have limited eMMC storage or small SSDs, and their operating systems are designed to manage resources efficiently. A browser that demands dozens of gigabytes may be a poor fit for such hardware, especially when cloud-based AI features could handle the same tasks with less local impact.
Users who are concerned about storage and bandwidth can take comfort in the fact that Chrome's standard browsing experience remains fully functional without on-device AI. Disabling the features does not affect ordinary web rendering, extensions, or other browser tools. The only thing lost is the convenience of offline AI assistance. That trade-off might be very reasonable for people who do not need AI-powered writing help or automatic tab grouping.
As Google continues to refine its AI strategy, it will be interesting to see whether the company adjusts the storage requirement to make it more accessible. A 20GB threshold seems designed for modern mid-range hardware, but it ignores the reality of many budget machines. For now, users who want to reclaim space should head to settings and turn off the on-device AI toggle, while those who value local processing can make the extra room and enjoy the convenience of AI features that always work, even without an internet connection.
Source:Android Authority News