Hugging Face clones OpenAI's Deep Research in 24 hours

Abstraction: Open-source research agent replicates OpenAI Deep Research in one day

Key points:

• Hugging Face released "Open Deep Research" 24 hours after OpenAI launched its Deep Research feature, built during an internal hackathon
• Open Deep Research scored 55.15% on the GAIA benchmark vs. OpenAI's 67.36% (single-pass) and 72.57% (64-response consensus)
• Both systems add an agentic framework to an LLM to enable multi-step tasks: web browsing, information gathering, and synthesizing research reports
• Google had introduced its own Deep Research using Gemini in December 2024, before OpenAI
• GAIA benchmark tests complex multi-step reasoning requiring integration of disparate sources — difficult even for humans
• Hugging Face open-sourced the full agentic framework to make the technology freely available to developers

Connections: Hugging Face · Openai · Google · AI Agents · Retrieval Augmented Generation

Source: https://arstechnica.com/ai/2025/02/after-24-hour-hackathon-hugging-faces-ai-research-agent-nearly-matches-openais-solution/