Coding

Chrome's AI features may be hogging 4GB of your computer storage

Google's Chrome browser is quietly consuming up to 4GB of user storage with its Gemini and Nano AI features, a development that could exacerbate storage constraints on low-end devices. The culprit lies in the browser's caching of AI models, which are stored in a proprietary binary format known as the "Gemini model cache." This hidden storage hog could have significant implications for Chrome users with limited storage capacity.

Daniel B (AI-assisted) May 10, 2026 1 min read EN

Based on reporting from Source.

Google Chrome's AI features, specifically Gemini and Nano, may be consuming up to 4GB of user storage due to the caching of AI models in a proprietary binary format known as the "Gemini model cache." This storage usage could exacerbate constraints on low-end devices.

Overview

The Gemini Nano AI model powers Chrome AI tools like scam detection, writing assistance, autofill, and suggestion features. As the model is designed to run locally, it needs to use training parameters stored on the user's device rather than pulling information from cloud-based models. This provides some privacy benefits but may not be ideal for users with limited storage.

What it does

The weights.bin file connected to the Gemini Nano AI model is automatically downloaded to the browser's system folders when certain AI features are enabled. Users can check for the file by opening their Chrome data folders and inspecting the OptGuideOnDeviceModel directory. However, simply deleting the file is not a solution, as Chrome may re-download it if AI features are still enabled.

Tradeoffs

To remove the file and prevent it from being re-downloaded, users need to head to Settings > System and toggle off the On-Device AI option. This will disable the AI features and prevent the file from coming back. Google specifies that the exact size of the Gemini Nano model may vary as the browser updates the model, but this information is not clearly presented to users when enabling AI features.

In a statement, Google spokesperson Scott Westover said, "We've offered Gemini Nano for Chrome since 2024 as a lightweight, on-device model. It powers important security capabilities like scam detection and developer APIs without sending your data to the cloud." Google has also begun rolling out the ability for users to easily turn off and remove the model directly in Chrome settings.

In conclusion, Google Chrome users with limited storage capacity should be aware of the potential storage usage of the Gemini and Nano AI features. To manage storage, users can disable the On-Device AI option in Chrome settings, which will remove the AI features and prevent the weights.bin file from being downloaded.

More articles like this

Coding 1 min

Visual Studio Code 1.120

Visual Studio Code’s 1.120 update slashes debugging friction with native Data Breakpoints, letting engineers pause execution when specific object properties change—not just memory addresses. The release also bakes in GitHub Copilot-powered inline code completions for Python, JavaScript, and TypeScript, cutting keystrokes by up to 40% in early benchmarks, while a revamped terminal shell integration finally bridges the gap between local and remote workflows.

Coding 1 min

Academic Research Skills for Claude Code

A new GitHub repository reveals the inner workings of Claude's code, exposing a critical gap in its knowledge graph: the lack of explicit academic research skills, which could compromise its ability to reason about complex, domain-specific topics, and instead relies on general knowledge and web scraping. The repository's author has identified a series of missing skills, including literature review and citation analysis, which are essential for informed decision-making in various fields. This oversight highlights the need for more nuanced knowledge representation in large language models.

Coding 1 min

Space Cadet Pinball on Linux

A long-dormant Linux port of the classic pinball game Space Cadet has been revived, leveraging the Escrow protocol to sidestep the original game's proprietary dependencies and enable seamless integration with modern Linux distributions. This update not only breathes new life into a nostalgic favorite but also underscores the potential of Escrow to facilitate the preservation and migration of legacy software. The reimagined game is now available for installation via standard package managers.

Coding 1 min

Gen Z Resentment Toward AI Grows as Adoption Stagnates and Workplace Fears Mount

As AI adoption stalls and workplace automation fears intensify, a growing Gen Z backlash against AI-powered tools is fueled by concerns over job displacement and lack of transparency in decision-making processes, with 71% of respondents citing algorithmic bias as a major concern. This sentiment shift is particularly pronounced among younger workers, who are increasingly wary of AI-driven hiring tools and performance evaluation systems.

Coding 1 min

Task Paralysis and AI

As AI systems increasingly rely on task decomposition, a phenomenon known as "task paralysis" is emerging, where overly complex task hierarchies hinder model performance and efficiency. This issue arises from the proliferation of nested, recursive task structures, which can lead to combinatorial explosion and decreased model scalability. Researchers are now exploring novel decomposition techniques, such as graph-based task representations, to mitigate this problem.

Coding 1 min

Gemini API File Search is now multimodal

Google's Gemini API file search capabilities have expanded to incorporate multimodal interaction, allowing developers to query and retrieve files via natural language prompts, images, and even audio inputs, leveraging a robust Relevance Aware Generator (RAG) model to improve search accuracy and efficiency. This upgrade enables more intuitive and flexible file search functionality, potentially revolutionizing workflows in industries reliant on large document archives. The API now supports a broader range of input modalities.