OpenAI Expands Access to Deep Research
(Image generated with AI on Shutterstock)
OpenAI announced it has expanded access to its Deep Research, its new AI agent designed to conduct complex research. Deep Research debuted earlier this month to users of ChatGPT Pro, the company’s $200/month subscription tier, but is now available to all paid users.
“Deep Research is now rolling out to all ChatGPT Plus, Team, Edu, and Enterprise users,” the company said in a social media post Tuesday.
Previously, Pro users were allowed up to 100 queries per month. Now, Pro users are allotted 120 queries per month, while Plus, Team, Edu, and Enterprise users get 10 queries per month.
OpenAI also announced improvements to Deep Research since its initial launch, including the addition of citations for embedded images and enhanced capabilities for understanding and referencing uploaded files.
“Give it a prompt and ChatGPT will find, analyze and synthesize hundreds of online sources to create a comprehensive report in tens of minutes vs what would take a human many hours,” the company said of the new Deep Research capability, accessible through the ChatGPT interface.
Deep Research is an AI agent designed for in-depth internet research, capable of conducting multi-step inquiries, analyzing text, images, and PDFs, and adapting its approach based on new information. Powered by an early version of OpenAI’s o3 model, it also supports user-provided files and can execute Python code for data analysis.
(Source: OpenAI)
OpenAI says Deep research is built for people who do intensive knowledge work in areas like finance, science, policy, and engineering and need thorough and reliable research. It's also useful for discerning shoppers looking for hyper-personalized recommendations on purchases that require careful research, the company says.
In a tweet, OpenAI CEO Sam Altman said, “Deep Research out for ChatGPT Plus users!” and called the feature “One of my favorite things we have ever shipped.”
OpenAI CPO Kevin Weil also said it was his favorite product the company has launched and described how he has used it: “It can do weeklong research-oriented tasks in 15 mins. I've used it to better understand muon colliders, the renewable energy market, and AI post training techniques—and to research/purchase a basketball hoop for my kids and new shoes for myself.”
Deep Research did well on a benchmark test called “Humanity’s Last Exam,” made by researchers at the Center for AI Safety in conjunction with data annotation firm Scale AI. The test is made up of roughly 3,000 multiple-choice and short answer questions submitted and reviewed by experts in academic fields like mathematics and philosophy.
Deep Research scored 26.6% on the test, the current leading score, which may not sound impressive at first glance. But the previous leaders, OpenAI’s o1 and DeepSeek’s R1 models, both only scored 9%. This represents a large jump in intelligence for Deep Research, with the largest gains occurring on questions related to chemistry, humanities and social sciences, and mathematics, according to OpenAI.
Still, all large language models are prone to hallucinations, or completely fabricating facts and events that can look and sound credible. Those using generative AI for research must carefully check model outputs for accuracy, despite how well the models do on benchmarks.
That is a reality recently faced by Florida-based injury law firm Morgan and Morgan, who was fined by a Wyoming judge for submitting a case filing with eight fictitious court cases generated by a chatbot.
“This deeply regrettable filing serves as a hard lesson for me and our firm as we enter a world in which artificial intelligence becomes more intertwined with everyday practice,” the case’s lead attorney Mike Morgan wrote in a declaratory filing, adding that “there are no shortcuts in law.”
