OpenAI has unveiled Deep Research, a new autonomous research capability for ChatGPT that can analyze hundreds of online sources and synthesize comprehensive reports in a fraction of the time it would take humans.
The feature, which launches today for ChatGPT Pro subscribers, represents a significant advancement in AI's ability to conduct complex, multi-step research tasks.
Deep Research differentiates itself by showing users its complete research process through a sidebar that displays steps taken and sources consulted. The agent can analyze text, images, and uploaded files like PDFs or spreadsheets to provide context-aware responses.
According to OpenAI, the system takes 5 to 30 minutes to complete research tasks, delivering results at the level of a professional research analyst.
The technology is powered by a version of OpenAI's upcoming o3 model, specifically optimized for web browsing and data analysis.
In benchmark testing, the model achieved 26.6% accuracy on "Humanity's Last Exam," a comprehensive evaluation covering over 3,000 expert-level questions across 100+ subjects. This score significantly outperforms previous models, including GPT-4o's 3.3% and other leading AI systems.
Due to its compute-intensive nature, Deep Research access will be rolled out gradually. ChatGPT Pro users ($200/month) will initially receive up to 100 queries per month, with plans to extend "limited access" to Plus, Team, and Enterprise users later.
OpenAI has announced that all paid users will eventually receive higher rate limits when a more cost-effective version becomes available.
The company acknowledges certain limitations, including potential hallucinations, challenges in distinguishing between authoritative information and rumors, and occasional formatting errors in reports and citations. OpenAI expects these issues to improve with increased usage and continued development.
Looking ahead, OpenAI plans to expand Deep Research's capabilities by enabling connections to specialized data sources, including subscription-based and internal resources.
The feature will also be integrated into mobile and desktop apps within the month, marking another step toward OpenAI's vision of creating more sophisticated AI agents capable of autonomous research and execution.