Scientists praise OpenAI's Deep Research, its new AI agent that searches internet more thoroughly

OpenAI, the AI wizz behind ChatGPT, has just launched a new agent. It’s called a “Deep Research” mode, which lets you do multi-step research on the internet for complex tasks.

Sam Altman, OpenAI’s boss, says that the Deep Research mode gives you expert-level reports in moments on tasks that you normally spend hours or even days to finish, not to mention a big chunk of money for it.

Here’s how it works:

While it is currently compute-intensive and slow (up to 30 minutes per task) and still prone to hallucination, it is capable of handling a wide variety of valuable tasks.

“50 cents of compute for 500 dollars of value,” Altman says.

Derya Unutmaz, a biomedical scientist over at The Jackson Lab, says that the “game-changer” of an agent can help with tasks on “scientific research, publishing, legal documents, medicine, education” and “likely many others.” The agent itself was built on OpenAI’s o3, its latest reasoning-excelling model.

The Deep Search agent is now available in the $200/month Pro tier with 100 queries per month, and future access will expand to the $20/month Plus, Team, and Enterprise tiers, eventually reaching the free tier.

OpenAI has been doing a lot more than making a fast AI chatbot: the models could actually exceed humans’ computing capability, whether we’re open to it or not.

More recently, the Microsoft-backed company partnered with the US National Laboratories for its latest reasoning models, including the o1 and potentially the o3, on the Venado supercomputer at Los Alamos National Laboratory.