After 24-hour hackathon, Hugging Face’s AI research agent nearly matches OpenAI’s solution

Hugging Face researchers developed "Open Deep Research," an open source AI research agent, to rival OpenAI's similar feature. After 24 hours, it achieved 55.15% accuracy on the GAIA benchmark, compared to OpenAI's 67.36%. The agent framework significantly enhances AI model capability, with GPT-4 alone scoring 29% on GAIA versus 67% with the framework. Hugging Face's solution utilizes "code agents" and existing open source tools, enabling quick improvements and community input.