Following the launch of Operator last week, OpenAI has introduced its latest AI agent, Deep Research. Designed for in-depth research, it utilizes ChatGPT along with OpenAI o3, optimized for web browsing and Python analysis. Deep Research applies advanced reasoning to explore text, images, and PDFs across the internet extensively, according to OpenAI.
In a blog post, OpenAI stated, “Deep Research is built for professionals engaged in intensive knowledge work in fields like finance, science, policy, and engineering, requiring thorough and reliable research.”
“It’s also valuable for discerning shoppers seeking hyper-personalized recommendations for purchases that require careful research,” the post adds.
OpenAI VP of Engineering, Srinivas Narayanan, describes the new model as “a new way of using AI to assist with complex knowledge tasks.” He compares Deep Research to consulting an expert, highlighting its ability to streamline tasks and save hours of manual research.
How to Use OpenAI’s Deep Research
In ChatGPT, select “Deep Research” from the message composer and enter your query. Whether you’re looking for a competitive analysis of streaming platforms or a personalized report on the best commuter bike, simply ask ChatGPT. You can also attach files or spreadsheets for additional context. After submitting your request, a sidebar will display the steps taken and the sources referenced throughout the process.
OpenAI states that Deep Research can take anywhere from 5 to 3o minutes, depending on the complexity of the query. You will receive a notification once the research is complete and can continue with other tasks in the meantime. The final output is delivered as a detailed report within the chat.
While Deep Research is designed to provide in-depth, well-documented answers with verified sources, OpenAI notes that GPT-4 is better suited for real-time, multimodal conversations. For complex, domain-specific queries requiring thorough exploration and precise citations, Deep Research stands out, delivering reliable and actionable insights.
CEO Sam Altman shared an example on X to illustrate how this AI agent works. He acknowledged that while the model operates slowly, it is the first AI capable of handling a broad range of complex and valuable tasks.
What’s Next?
OpenAI plans to introduce embedded images, data visualizations, and other analytical features to enhance clarity and provide deeper context in reports.
How is it Different from Operator?
The key difference lies in their functionalities, while the Operator handles different repetitive browser tasks like filling out forms, ordering groceries, or even generating memes, it lacks the capability for in-depth research.
Deep Research, on the other hand, builds upon advanced reasoning to tackle complex problems in both professional and everyday scenarios.
Despite their differences, both AI agents mark a significant step toward OpenAI’s broader vision of achieving AGI (artificial general intelligence), a goal the company has long emphasized.
This development also comes amid rising attention on the Chinese AI model DeepSeek, which has gained traction for its advanced technology and affordability. OpenAI CEO Sam Altman acknowledged DeepSeek’s progress but promised even more powerful models in the future.