OpenAI unveiled its latest innovation, the OpenAI Deep Research tool, during a high-profile event on Sunday. This tool is designed to crawl the web and compile comprehensive research reports on nearly any subject. However, access to this advanced utility is limited to users subscribed to OpenAI's premium ChatGPT Pro plan, which costs $200 per month. At the heart of this tool lies the proprietary o1 model, which is accessible only through a paid API.
In response to this development, a group of developers from the AI development platform Hugging Face have introduced an "open" version of this research tool, aptly named Open Deep Research. This initiative underscores Hugging Face's commitment to making advanced AI tools more accessible. Thomas Wolf, co-founder and chief scientist of Hugging Face, described their project as an "open" version of OpenAI's deep research tool.
Open Deep Research combines an AI model with an open-source "agentic framework" that aids the AI in planning its analysis and directs it to utilize various tools like search engines. Despite its promising approach, Open Deep Research achieved a score of 54% on the GAIA benchmark for general AI assistants, compared to OpenAI's Deep Research, which scored 67.36%.
OpenAI's tool not only outperforms open models like DeepSeek's R1 but also can read files across the web using a simple text-based browser paired with a "text inspector" toolkit. Remarkably, OpenAI achieved these capabilities by harnessing the o1 model in less than 24 hours. Despite its cutting-edge technology, users have reported that the tool is currently under heavy load, often resulting in error messages after 10 minutes of use.
The launch of OpenAI Deep Research has sparked several attempts at reproducing its functionality using open models and tools available online. These reproductions reflect a growing interest in democratizing access to advanced AI research capabilities.
Leave a Reply