ChatGPT Can Now Complete a Major Task That Would Take a Human Up to 30 Days. Here's How it Works.

ChatGPT tin now behaviour investigation and constitute a study arsenic thoroughly arsenic a investigation analyst, OpenAI claims.

OpenAI launched a caller AI supplier capacity connected Sunday called Deep Research, which infuses ChatGPT pinch the powerfulness to hunt the web, analyse sources, and find if those sources are relevant. Based connected its findings, Deep Research produces a broad investigation insubstantial pinch afloat citations that tin sometimes tally longer than 10,000 words.

"It tin do analyzable investigation tasks that mightiness return a personification anyplace from 30 minutes to 30 days," OpenAI's main merchandise serviceman Kevin Weil stated astatine an arena successful Washington showcasing the exertion past week. Deep Research tin get the aforesaid tasks done successful 5 to 30 minutes, he said.

While AI chatbots provide conversational responses to personification queries, AI agents spell a measurement further by completing multi-step tasks connected their own. Deep Research is an AI supplier that tin surf the Internet and synthesize accusation from various sources, and it is the 2nd AI supplier from OpenAI this year. OpenAI introduced Operator past month, an AI supplier that tin shop for groceries online aliases make a edifice reservation.

Related: OpenAI Just Released Its Text-to-Video Generator, Sora. Here's How the New AI Could Impact Small Businesses and Creators.

Deep Research has already hit retired the competition, including DeepSeek's R1, xAI's Grok 2, and Google's Gemini Thinking, connected a caller information called "Humanity's Last Exam" released past period by researchers astatine the Center for AI Safety and Scale AI.

Researchers declare that the 3,000-question exam is the astir difficult trial ever fixed to AI systems. While DeepSeek's R1 scored 9.1% connected the test, Google's Gemini Thinking scored 6.2%, and Grok 2 scored 3.8%, Deep Research scored 26.6%.

How does Deep Research work?

In a demo introduction to Deep Research released connected YouTube connected Sunday, OpenAI merchandise task head Neel Ajjarapu demonstrated really Deep Research was accessible arsenic a fastener wrong the hunt barroom of ChatGPT.

Accessing Deep Research is arsenic elemental arsenic pressing a fastener connected the homepage of ChatGPT.

Ajjarapu wanted Deep Research to look into whether OpenAI should create a caller connection translator app and typed successful a circumstantial query: "Help maine find iOS and Android take rates, % who want to study different language, and alteration successful mobile penetration, complete the past 5 years, for apical 10 developed and apical 10 processing countries by GDP. Lay this info retired successful a formatted report, a array connected metrics, and see recommendations connected markets to target for a caller translator app from ChatGPT, focusing connected markets ChatGPT could amended grow to."

Deep Research responded to Ajjarapu's punctual pinch a bid of questions, including if he wanted wide iOS and Android take rates, aliases circumstantial categories, for illustration instrumentality shipments.

After Ajjarapu answered the clarifying questions, Deep Research was disconnected to work. The AI supplier kept Ajjarapu updated connected what it was doing successful ChatGPT, for illustration searching for the apical 10 countries by GDP.

While DeepResearch goes to work, it keeps users updated connected what it's doing successful the sidebar to the right. Credit: OpenAI

Meanwhile, Deep Research was surfing the web connected its own, knowing text, images, and PDFs, and synthesizing information. Ajjarapu said that Deep Research could beryllium utilized for marketplace research, world investigation crossed departments, and astatine work.

In 11 minutes, Deep Research looked done 29 websites in-depth to create a investigation study complete pinch tables, slug points, and cited sources. This study would person taken "hours" to put together, Ajjarapu stated.

Related: Would You Pay $200 for ChatGPT? OpenAI's New Reasoning Model Has a Hefty Price Tag.

Deep Research is now embedded wrong ChatGPT for OpenAI Pro users, pinch plans to rotation it retired to paying Plus and Team users next.