OpenAI ChatGPT Desktop App Bundles 1.7GB of Runtimes, Including LibreOffice
Developer Simon Willison discovered the app installs a headless office suite and multiple programming environments in the user's cache.
The OpenAI ChatGPT desktop app, formerly known as Codex, is bundling a massive suite of third-party software and runtimes directly into user cache folders. The discovery highlights the significant local infrastructure OpenAI is deploying to support advanced document processing and code execution.
Developer Simon Willison found that the app's `codex-primary-runtime` directory on macOS contains a full installation of several heavy-duty binaries. Most notably, the app bundles a "libreoffice-headless" version, which occupies approximately 429.7 MB of disk space. Beyond the office suite, the runtime folder includes Python (440.6 MB), Node.js (446.4 MB), Poppler (187.9 MB), and git (148.1 MB). In total, the `codex-primary-runtime` folder consumes approximately 1.7 GB of storage, located at `~/.cache/codex-runtimes/codex-primary-runtime/dependencies/native`.
The Drive for Local Consistency
OpenAI has been steadily evolving the ChatGPT desktop experience to integrate deeper local capabilities for document analysis and code execution. By bundling these runtimes, OpenAI ensures that the AI's tools behave consistently across different user environments. This approach removes the need for users to manually install complex dependencies or manage version conflicts on their own systems, which would otherwise be a significant barrier to the app's functionality.
Implications for Disk Footprint
This distribution strategy allows the AI to perform high-fidelity document conversion and analysis locally or within a controlled environment, leveraging LibreOffice headless for file processing without relying on the user's system configuration. However, the trade-off is a substantial increase in the app's disk footprint. Bundling an entire office suite alongside multiple language runtimes raises questions about software distribution efficiency and the long-term management of these dependencies as the app grows.
What to Watch
While the presence of these binaries is confirmed, the exact internal mechanism the app uses to trigger these tools remains a point of interest. It is currently unclear how OpenAI plans to update these bundled runtimes to address security vulnerabilities or version obsolescence without requiring massive downloads for the end user. As AI agents move toward more autonomous local operation, the industry may see a trend toward these "heavy" bundles to ensure reliability at the cost of storage.