OpenAI has introduced a new ChatGPT feature called “deep research”, which is designed to gather and analyze information from the internet.
The feature is aimed at users who need detailed and accurate reports, particularly in fields like finance, science, policy, and engineering. It can also help with research before making major purchases, such as cars or home appliances.
According to OpenAI's blog post published on February 2, the company explains that the feature provides “fully documented” results, complete with citations and a summary of its reasoning. Depending on the request, it may take anywhere from five to thirty minutes to generate a response.
Did you know?
Subscribe - We publish new crypto explainer videos every week!
What is a Crypto Bridge? (Explained with Animations)
“Deep research” runs on OpenAI’s latest model, OpenAI o3, which is designed to browse the web and analyze data more effectively. The model aims to improve accuracy by checking its own findings. However, OpenAI acknowledges that this new feature may still produce incorrect details or struggle to separate reliable sources from misleading ones.
The new agent recently performed well on a test called Humanity’s Last Exam, which consists of 3,000 expert-level questions across more than 100 topics. It achieved a 26.6% accuracy rate, higher than DeepSeek-R-1’s 9.4% and GPT-4o’s 3.3%.
This release follows OpenAI’s launch of another ChatGPT tool, Operator, on January 23. That feature was built for completing online tasks like booking travel or ordering groceries.
Recently, OpenAI and Microsoft investigated whether a group linked to Chinese AI company DeepSeek used OpenAI's data without permission. How did it go? Read the full story.
