OpenAI Unveils Deep Research, a New AI Agent That Can Complete Multi-Step Research Tasks

OpenAI has officially launched Deep Research, a new AI agent that uses advanced o3 model reasoning to synthesize information from massive amounts of data on the web. ChatGPT users will be able to use this feature to quickly complete multi-step research tasks. ChatGPT Pro subscribers will be able to experience this feature starting today with a limit of 100 queries per month. Deep Research is also planned to be rolled out to ChatGPT Plus and Team users soon, while free users will get access later.

Researchers typically spend hours searching, analyzing, and synthesizing hundreds of online sources to produce a standard research report. With ChatGPT Deep Research, the entire task can now be completed in just minutes. The new feature is trained on real-world tasks, using a browser and Python tools. The final report from Deep Research will include clear citations and summaries of the AI's thought process. OpenAI says the feature will be especially effective at finding in-depth, non-obvious information that often requires browsing multiple websites.

OpenAI Unveils Deep Research, a New AI Agent That Can Complete Multi-Step Research Tasks Picture 1

How to use Deep Research feature on ChatGPT:

1. Access ChatGPT on the web. (This feature will be available on mobile and desktop apps later this month.)
2. In ChatGPT, after entering the query, select Deep Research in the editor.
3. Attach a file or spreadsheet to provide additional context to the question.
4. As the process begins, a sidebar will display a summary of the steps taken and the information sources used.
5. Deep Research can take anywhere from 5 to 30 minutes to complete, depending on the complexity of the query.
6. When the report is ready, the user will receive a notification.

Currently, the final results are displayed as a text report within the chat. However, in the coming weeks, ChatGPT will add visualizations, data charts, and other analytics to the report to provide more clarity and detail. In the future, ChatGPT will also allow users to connect to more in-depth data sources to increase accuracy and personalize results.

Notably, Deep Research is said to have performed impressively in AI tests. Humanity's Last Exam is a new assessment that tests AI models across a variety of domains using expert-level questions. The exam consists of more than 3,000 multiple-choice and essay questions spanning more than 100 topics. OpenAI's Deep Research model scored a record 26.6% on the test. OpenAI also claims that the model has the highest score in GAIA, another public benchmark for AI on real-world questions.

With its ability to aggregate information and provide detailed citations, Deep Research promises to significantly save time on research tasks. This will certainly be a useful tool for many users. You can learn more about OpenAI's Deep Research feature here.

Jessica Tanner

Update 06 February 2025