The artificial intelligence development platform Hugging Face recently announced that they have successfully built an "open" version of OpenAI's deep research tool, aimed at competing with OpenAI's recently released deep research tools.
OpenAI showcased its deep research tool at an event, which can automatically compile research reports on various topics by scraping information from the internet. Unfortunately, this tool currently offers limited access only to users subscribed to OpenAI's $200 per month ChatGPT Pro plan.
The team at Hugging Face, including co-founder and chief scientist Thomas Wolf, stated that their open deep research project combines OpenAI's o1 model with an open-source "agent framework." This framework is designed to help the model analyze information better and guide its use of tools like search engines. Although o1 is a paid proprietary model, researchers believe it outperforms some open models, such as DeepSeek's R1.
In less than 24 hours, the research team successfully developed a simple text browser and "text checker" toolkit using the o1 model, allowing open deep research to navigate the web autonomously. This tool can scroll through web pages, manipulate files, and even perform data calculations, thereby enhancing information retrieval efficiency.
According to the GAIA benchmark test, the score for open deep research is 54%. In contrast, OpenAI's deep research scored 67.36%. Nevertheless, when testing open deep research, a journalist attempted to access the public demo page set up by the Hugging Face team but was unable to use it due to high traffic, ultimately receiving an error message.
The research team has committed to improving the user experience and has uploaded the source code to GitHub for public viewing and feedback. It is worth noting that several "replicas" of OpenAI's deep research have already appeared online, but most of these versions rely on open models and tools, lacking the crucial o3 model. In benchmark tests for complex problem-solving and information retrieval, the o3 model's performance is nearly unmatched.
As the market demand for deep research tools continues to grow, it remains to be seen whether Hugging Face's new endeavor can compete with OpenAI's products in the future, and this is something we should keep an eye on.
Project: https://github.com/huggingface/smolagents/tree/gaia-submission-r1/examples/open_deep_research
Key Points:
🚀 Hugging Face launches an "open" version of OpenAI's deep research tool, aimed at competing with OpenAI's deep research tools.
🌐 This tool combines the AI model o1 with an open-source framework, enabling autonomous web information retrieval.
💻 The research team has released the source code on GitHub, committed to improving user experience and welcoming feedback.
