New ChatGPT Agent for Advanced Research
OpenAI's new ChatGPT agent takes research to the next level by searching, analyzing, and compiling information from hundreds of sources online. With the ability to draw its own conclusions and deliver detailed reports, the AI agent could save professionals in finance, tech, and
OpenAI is now introducing "Deep Research" in ChatGPT - an AI agent that can carry out multi-step research assignments on the internet, analyze large amounts of data, and compile detailed reports. The feature is available to Pro users starting today and will soon also be offered to Plus and Team users.
What Is Deep Research?
Deep Research is a new agent feature in ChatGPT that can carry out comprehensive analysis by searching, interpreting, and compiling information from hundreds of sources online. It uses an advanced version of OpenAI's upcoming o3 model, optimized for web search and data analysis.
The AI agent functions as a digital research assistant and can quickly produce detailed reports on things like competitive analysis, market research, and technical evaluations. It can also handle user-uploaded files, analyze tables, and include visualizations in its results.
An AI That Thinks and Summarizes on Its Own
Deep Research is built on advanced reasoning capability, meaning it can not only gather information but also draw insightful conclusions. It can:
Search for relevant sources and identify credible information
Compile and analyze data to create a detailed report
Citations and source references for transparency and verifiability
Who Benefits From Deep Research?
Deep Research is primarily aimed at professionals in fields like finance, science, policy, and engineering, where the need for in-depth, reliable information is crucial. The AI agent can also be a great help for individuals who want to do extensive product comparisons or other complex research.
How Deep Research Is Used
To activate Deep Research in ChatGPT, the user selects the feature in the message field and types in a question. During the analysis, a side panel shows the AI agent's steps and sources. A finished report is delivered after 5-30 minutes, depending on the complexity of the task.
OpenAI will soon integrate more features, such as embedded images, interactive data visualizations, and more advanced information processing.
Deep Research Compared to GPT-4o
GPT-4o is optimized for real-time conversations and multimodal interaction, while Deep Research is built for more in-depth, complex analysis. The AI agent's ability to investigate, verify, and structure information makes it especially valuable for advanced research assignments where accuracy and source citation are crucial.
Groundbreaking Results in AI Tests
The model powering Deep Research has performed exceptionally well in tests like "Humanity's Last Exam" and GAIA, where AI is evaluated on its ability to solve expert-level tasks. In these tests, OpenAI's Deep Research model has outperformed previous top models, especially in areas like chemistry, math, and social science.
Limitations and Future Plans
Despite its high capability, Deep Research has certain limitations. It can still make incorrect interpretations, have difficulty determining the credibility of certain sources, and sometimes present facts with excessive confidence. OpenAI is, however, actively working to improve the AI agent's ability to handle uncertainty and to streamline its format and speed.
Deep Research is currently available to Pro users with a limit of up to 100 requests per month. Plus and Team users will soon get access, followed by a wider rollout. OpenAI also plans to make the AI agent faster and more cost-effective through a smaller model that can deliver high-quality results.
In the future, Deep Research will be integrated with more external databases and subscription services, making it even more powerful for professional users.
A New Era for AI-Powered Research Tools
With Deep Research, OpenAI is taking another step toward creating a truly autonomous AI assistant for knowledge-intensive tasks. By combining it with other AI features, like OpenAI's "Operator" for task execution, ChatGPT could soon become a powerful platform for both research and practical decision-making.