Exciting news for AI enthusiasts! A team at Hugging Face, led by co-founder and chief scientist Thomas Wolf, has developed an open-source version of OpenAI’s deep research tool. This new project, dubbed Open Deep Research, aims to make advanced AI-driven web research more widely available.

OpenAI’s deep research tool, which scours the web to compile detailed reports on any topic, is currently limited to users with a $200-a-month ChatGPT Pro subscription. Hugging Face’s Open Deep Research breaks down this barrier by combining an AI model (O1) with an open-source framework that guides the model to use tools like search engines and text analyzers.

What makes this project stand out is its speed and efficiency. In under 24 hours, the team was able to train the O1 model to autonomously navigate the web, scroll through pages, manipulate files, and even perform data calculations. On the GAIA benchmark for AI assistants, Open Deep Research scored 54%, compared to OpenAI’s 67.36%.

While the public demo faced some technical hiccups during testing, the researchers remain committed to refining the tool and have already shared the source code on GitHub for feedback.

It’s worth noting that several other “reproductions” of OpenAI’s deep research tool exist online. However, most of these lack O3, the proprietary model that powers OpenAI’s tool and outperforms others in complex question-answering and information-gathering tasks.

For now, Open Deep Research is a promising step toward making advanced AI tools more accessible, even if it’s not quite on par with OpenAI’s offering. Stay tuned for updates as the Hugging Face team continues to improve their open-source project!