Crispy Bites: OpenAI has launched “Deep Research,” an AI Agent designed to perform complex, multi-step internet research tasks. This tool can simulate a human research analyst, accomplishing in ten minutes what might take a human several hours.
Breakdown:
Performance: In the challenging “Humanity’s Last Exam,” Deep Research achieved a 26.6% accuracy rate, significantly outperforming competitors like ChatGPT o3-mini and DeepSeek’s R1 V3-powered model.
Research Capabilities: Deep Research excels in web browsing and data analysis, autonomously planning and executing research tasks. Users can input queries via text, images, and files, and the AI generates comprehensive responses within 5 to 30 minutes, providing a summary of its process and citations.
Availability and Access: Currently, Deep Research is available to ChatGPT Pro subscribers in the U.S. for a $200 monthly fee, with a limit of 100 queries per month. There are plans to expand access to other paid plans in the future.
Limitations: Despite its advanced capabilities, Deep Research may still produce inaccuracies, struggle to distinguish authoritative information from rumors, and occasionally generate incorrect inferences.
