OpenAI’s new ChatGPT agent is ‘like a superpower,’ says CEO Sam Altman

OpenAI has just announced a new AI tool called Deep Research.

The new AI agent is capable of conducting multi-step research on the internet for complex tasks and, according to OpenAI, “accomplishes in tens of minutes what would take a human many hours.”

You simply give it a prompt and ChatGPT will “find, analyze, and synthesize hundreds of online sources to create a comprehensive report at the level of a research analyst.”

OpenAI demonstrated the new feature in a video shared on Sunday (above).

Deep Research caters to professionals in finance, science, policy, and engineering, providing thorough, reliable insights. It’s also beneficial for shoppers seeking personalized recommendations on purchases that require careful research, like those for cars, appliances, and furniture. Outputs include clear citations and summaries, enabling easy verification. Essentially, the feature streamlines time-consuming research, delivering niche information efficiently from a single query.

In a flurry of posts shared on X on Sunday, OpenAI chief Sam Altman described Deep Research as being “like a superpower; experts on demand.”

He said it can “go use the internet, do complex research and reasoning, and give you back a report,” taking care of tasks “that would take hours/days and cost hundreds of dollars.”

While it’s “very compute-intensive and slow,” he claimed that “it’s the first AI system that can do such a wide variety of complex, valuable tasks.”

Deep Research takes between 5 and 30 minutes to complete its work, and you’ll receive a notification once the research is complete. The final output arrives as a report, delivered by ChatGPT. At the moment, reports are text-only, but OpenAI said that in the coming weeks it’ll be adding embedded images, data visualizations, and other analytic outputs for additional clarity and context.

Deep Research is now available as part of OpenAI’s Pro tier ($200 a month), with 100 queries available per month on the web (coming to mobile and desktop apps by the end of February). It’ll also arrive for Plus, Team, and Enterprise customers “soon,” before eventually making its way to OpenAI’s free tier.

Altman urged people to “give it a try on your hardest work task that can be solved just by using the internet and see what happens.”

But take note: OpenAI cautioned the its new tool “may struggle with distinguishing authoritative information from rumors, and currently shows weakness in confidence calibration, often failing to convey uncertainty accurately.” It added that, at launch, you might also see “minor” formatting errors in reports and citations, and tasks may take longer to kick off. “We expect all these issues to quickly improve with more usage and time,” OpenAI said.

Deep Research arrives just a week after OpenAI unveiled another AI agent, called Operator, which acts like a human assistant, performing web-based tasks like making reservations, booking trips, and ordering groceries, in line with your requests.