Hugging Face has developed an open-source version of autonomous research technology, matching key capabilities of OpenAI’s recently launched Deep Research feature. As reported by Benj Edwards for Ars Technica, the project called “Open Deep Research” was completed within 24 hours of OpenAI’s announcement. The new tool enables AI models to independently browse the web and create research reports.
The system achieved 55.15 percent accuracy on the General AI Assistants benchmark, compared to OpenAI’s 67.36 percent. It uses Hugging Face’s “smolagents” library and code-based agents, which are reportedly 30 percent more efficient than JSON-based alternatives.
Project leader Aymeric Roucher explained that while the system currently relies on OpenAI’s language models through an API, it can be adapted to work with open-weights AI models. The framework significantly improves AI performance, as demonstrated by the difference between GPT-4o alone (29 percent) and with the agent framework (67 percent) on benchmark tests.
The project builds on existing technologies, including web browsing tools from Microsoft Research’s Magnetic-One agent project. Hugging Face has published the code on GitHub and is actively seeking contributors to enhance the system’s capabilities.
While the performance doesn’t yet match OpenAI’s solution, the open-source nature of the project allows developers to study and modify the technology freely. Future improvements may include support for additional file formats and vision-based web browsing capabilities.