Chrome’s 4GB AI model isn’t new, but you’re not wrong for being confused

Chrome 的 4GB AI 模型并非新事物，但你感到困惑也情有可原

All of Google’s products have been getting more AI features, including Chrome, which now offers split-screen Gemini chatbot support, the ability to automate web browsing, and more. Some desktop Chrome users have also noted that the browser appears to suddenly want more storage space for AI. This is true—Chrome does download a 4GB AI model for on-device processing. It’s been doing that for years, though. Google 的所有产品都在增加更多 AI 功能，Chrome 也不例外，它现在提供了分屏 Gemini 聊天机器人支持、网页浏览自动化等功能。一些桌面版 Chrome 用户注意到，浏览器似乎突然需要更多的存储空间来运行 AI。事实确实如此——Chrome 确实会下载一个 4GB 的 AI 模型用于设备端处理。不过，这种情况已经持续多年了。

Google hasn’t actually changed anything about Chrome’s on-device AI, but the confusion is understandable, as the company has done a poor job of explaining what it’s doing and why. This is, unfortunately, par for the course with Google’s AI efforts. Just this week, someone noticed that Chrome had downloaded a 4GB Gemini Nano model and inferred from its sudden appearance that Google was deploying that AI on all Chrome installs right now. That’s not exactly true. Google 实际上并没有对 Chrome 的设备端 AI 做任何改动，但用户的困惑是可以理解的，因为该公司在解释其行为和动机方面做得并不好。遗憾的是，这在 Google 的 AI 推广工作中已是常态。就在本周，有人发现 Chrome 下载了一个 4GB 的 Gemini Nano 模型，并据此推断 Google 正在向所有 Chrome 安装包中部署该 AI。这并不完全准确。

Google announced in 2024 that it would begin adding local AI capabilities to Chrome, powering features like Help Me Write, tab organization, and scam detection. The list of potential uses has changed over time, and Google never rolled the models out universally. It’s possible your machine has been running a local Gemini Nano model since 2024, and it’s also possible the AI was downloaded yesterday. Google 在 2024 年宣布将开始为 Chrome 添加本地 AI 功能，以支持“帮我写作”（Help Me Write）、标签页整理和诈骗检测等功能。潜在用途的列表随时间而变化，且 Google 从未在全球范围内统一推送这些模型。你的电脑可能从 2024 年起就在运行本地 Gemini Nano 模型，也可能是在昨天才下载的。

Google tells Ars Technica the flags that determine whether or not Gemini Nano is installed on your machine are multifaceted and include (but are not limited to) your computer’s hardware, account features, and whether you’ve visited a website that uses Google’s on-device Gemini API. Because Gemini Nano is constantly appearing on machines for the first time, people may think this is something new. In reality, Google confirmed to Ars that Chrome’s Gemini Nano model has been around 4GB in size since it debuted two years ago. Google 向 Ars Technica 表示，决定你的机器是否安装 Gemini Nano 的标志是多方面的，包括（但不限于）你的电脑硬件、账户功能，以及你是否访问过使用 Google 设备端 Gemini API 的网站。由于 Gemini Nano 不断在不同机器上首次出现，人们可能会认为这是新事物。实际上，Google 向 Ars 证实，Chrome 的 Gemini Nano 模型自两年前推出以来，大小一直约为 4GB。

The default problem

默认设置的问题

So Google has not suddenly decided to gobble up 4GB of your storage space—it did that two years ago, and nobody thought to get upset. But should you? A fresh Chrome install with no extensions eats up 6–8GB of storage, and it will grow considerably with cache and extension data. In a few months, Chrome can occupy 10 times more space than the AI model consumes. Maybe that’s not ideal, either, but that’s just the state of browsers today. 所以，Google 并不是突然决定占用你 4GB 的存储空间——它两年前就这么做了，当时没人觉得有什么问题。但你现在应该感到困扰吗？一个没有任何扩展程序的全新 Chrome 安装包会占用 6-8GB 的存储空间，随着缓存和扩展数据的增加，这个数字还会大幅增长。几个月后，Chrome 占用的空间可能是该 AI 模型所占空间的 10 倍。这或许并不理想，但这就是当今浏览器的现状。

Google says Chrome will delete the model if your machine runs low on storage, and you can turn off local AI features in the Chrome settings with a toggle under the System tab. Flip that off, and Chrome will remove the model and not redownload it. This is a fixable problem, but it shouldn’t be a problem in the first place. Google 表示，如果你的机器存储空间不足，Chrome 会删除该模型；你也可以在 Chrome 设置的“系统”选项卡下通过开关关闭本地 AI 功能。将其关闭后，Chrome 会移除该模型且不会重新下载。这是一个可以解决的问题，但它本不应该成为一个问题。

Running AI models locally is generally a good thing—that data stays on your machine rather than feeding a cloud-based model. It’s a positive for privacy and security, and users should have that choice. But “choice” is the operative word, and Google isn’t actually giving you one because AI is Google’s new default. Maybe you don’t want any of these AI features. If that’s the case, Chrome should not consume 4GB of space for something you don’t use without asking. 在本地运行 AI 模型通常是一件好事——数据保留在你的机器上，而不是用来喂养云端模型。这对隐私和安全是有利的，用户理应拥有这种选择权。但“选择”是关键词，而 Google 实际上并没有给你选择权，因为 AI 已成为 Google 的新默认设置。也许你根本不需要这些 AI 功能。如果是这样，Chrome 不应该在未经询问的情况下，为了你不需要的东西占用 4GB 空间。

A curious omission

一个令人好奇的遗漏

Google users were more willing to excuse AI in 2024, but the backlash is real in 2026. People are increasingly looking to avoid AI features, which makes this 4GB stealth download all the more questionable. Google’s obsession with AI has led to numerous stumbles, even when the company has ostensibly good intentions, because we are all (rightly!) hyper-focused on how this technology is impacting our lives. Google 用户在 2024 年更愿意包容 AI，但在 2026 年，抵制情绪已非常真实。人们越来越倾向于避开 AI 功能，这使得这次 4GB 的“静默下载”更显可疑。Google 对 AI 的痴迷导致了多次失误，即使该公司表面上怀有善意，因为我们所有人（理所当然地！）都高度关注这项技术如何影响我们的生活。

Some of those “good intentions” seem to have made the Chrome situation worse. As users sought ways to remove this AI model, many looked for the settings toggle. This happened to coincide with the wide release of Chrome 148, and the label for this toggle included a pretty suspicious change versus v147. Google claims this change was made to be clear about how Chrome’s APIs work. 其中一些“善意”似乎让 Chrome 的处境变得更糟。当用户寻找移除该 AI 模型的方法时，许多人找到了设置开关。这恰逢 Chrome 148 的广泛发布，该开关的标签与 v147 版本相比，包含了一个相当可疑的改动。Google 声称这一改动是为了更清楚地说明 Chrome 的 API 是如何工作的。

Google removed the stipulation that its on-device AI model would not send data to Google’s servers. This is alarming, as one of the primary benefits of local AI is its greater privacy. We reached out to Google to ask if this wording is due to a change in Chrome’s on-device AI. “This doesn’t reflect a change to how we handle on-device AI for Chrome,” a Google spokesperson said. “The data that is passed to the model is processed solely on device.” Google 删除了关于“其设备端 AI 模型不会向 Google 服务器发送数据”的规定。这令人担忧，因为本地 AI 的主要优势之一就是更高的隐私性。我们联系了 Google，询问这种措辞的改变是否是因为 Chrome 设备端 AI 的变动。一位 Google 发言人表示：“这并不反映我们处理 Chrome 设备端 AI 方式的改变。传递给模型的数据完全是在设备上处理的。”

According to Google, the team decided to make this change earlier in 2026 to ensure it was being crystal clear about how AI works on the web. Chrome’s local AI has an API that a site might use, for example, to do summarization or edit your writing. In these instances, the website would naturally see the input and output. If it’s a Google website, that data ends up on Google’s servers. If it’s a non-Google site, Google doesn’t see any of that data. 据 Google 称，团队在 2026 年初决定做出这一改动，以确保清楚地说明 AI 在网络上是如何工作的。Chrome 的本地 AI 提供了一个 API，网站可以使用它来进行摘要总结或编辑你的写作。在这种情况下，网站自然会看到输入和输出内容。如果是 Google 网站，这些数据最终会进入 Google 的服务器；如果是非 Google 网站，Google 则不会看到任何此类数据。

That explanation may or may not be satisfying as the backlash against AI grows. Regardless, using the web is never completely private. If you’re uncertain about using AI tools on a site, you should always try to parse its privacy policy, which will tell you how your data (AI-generated or not) will be used. As long as Google is deploying AI as an opt-out service, you’ll have to be extra vigilant. As the saying goes, it’s easier to ask for forgiveness than permission, but Google ought to be asking for permission a lot more. 随着对 AI 的抵制情绪日益高涨，这种解释是否令人满意尚不可知。无论如何，使用网络永远无法做到完全私密。如果你不确定是否要在某个网站上使用 AI 工具，你应该始终尝试阅读其隐私政策，它会告诉你你的数据（无论是否由 AI 生成）将如何被使用。只要 Google 将 AI 作为一种“默认开启、需手动关闭”的服务来部署，你就必须格外警惕。俗话说，请求原谅比请求许可更容易，但 Google 本应更多地请求用户的许可。