OpenAI Launches "Deep Research": An AI Tool to Compete with China's DeepSeek

A week after launching Operator, OpenAI has unveiled its new AI tool, Deep Research, designed for conducting detailed investigations using ChatGPT.

The new agent relies on OpenAI's o3 model, optimized for web browsing and data analysis with Python, allowing it to intelligently and systematically research texts, images, and PDF files across the internet.  

A Powerful Tool for Professionals and Consumers Alike

In a blog post, OpenAI explained that Deep Research is intended for individuals working in fields such as finance, science, policy, and engineering, who require reliable and comprehensive research. The company also noted that the feature would be beneficial for consumers seeking highly personalized recommendations before making complex purchasing decisions.  

Reducing Time and Effort with AI

Srinivas Narayanan, OpenAI’s Vice President of Engineering, stated that this technology represents "a new way to use AI to assist with complex cognitive tasks." He compared the new feature to consulting an expert, significantly reducing the time required for intensive research that would normally take hours of human effort.  

How Does "Deep Research" Work?

To use this feature in ChatGPT, users can select "Deep Research" from the menu and enter their query. Whether it's a competitive analysis of streaming platforms or a customized report on the best commuter bike, simply providing ChatGPT with the necessary details is enough. Users can also attach files or spreadsheets for additional context.  

Once the request is submitted, a sidebar displays the steps taken and sources used. The research process typically takes between 5 to 30 minutes, depending on the complexity of the topic, and users receive a notification upon completion. The final report is delivered within the chat in a detailed format with cited sources.  

"Deep Research" vs. "GPT-4" and "Operator" 

OpenAI emphasizes that Deep Research provides reliable, well-sourced answers, but it is slower compared to GPT-4, which is better suited for instant, multimodal conversations.  

For in-depth, data-driven analysis, Deep Research is the preferred choice, offering expanded exploration and precise citations, making it an ideal tool for gaining trustworthy and actionable insights.  

Sam Altman: "The First AI Model Capable of Handling Complex and Valuable Tasks with This Level of Diversity" 

OpenAI CEO Sam Altman shared a post on X (formerly Twitter) explaining how Deep Research works. He acknowledged its relatively slow speed but described it as a "major leap in AI’s ability to handle a wide range of complex tasks." OpenAI is currently working on adding integrated images, graphs, and analytical features to enhance the clarity and context of reports in the near future.  

How Is "Deep Research" Different from "Operator"?

The key difference is that Operator specializes in performing repetitive tasks across the browser, such as filling forms, ordering products, and generating memes, but it is not designed for deep research like Deep Research.  

On the other hand, Deep Research focuses on analysis and inference, making it better suited for advanced research tasks and complex problem-solving in both professional and everyday scenarios.  

OpenAI Faces Competition from China's DeepSeek  

This development comes at a time when China’s DeepSeek AI model has gained widespread attention due to its advanced technology and low cost. While Sam Altman has praised DeepSeek, he has also promised that OpenAI will continue to deliver stronger and more efficient models.