OpenAI has launched Deep Research, an AI agent designed for conducting autonomous research.

OpenAI has announced a new AI agent called Deep Research, designed for independently conducting research tasks assigned by the user. Once the topic and clarifying details are received, the agent begins its work, thoroughly describing each step of the process.

How does Deep Research work?

The AI agent develops a step-by-step plan, analyzes information, and adjusts its approach in real-time if necessary. This dynamic method allows the agent to revisit previous steps, refine data, and adapt its search.

The average time for completing a single research task ranges from 5 to 30 minutes. Upon finishing its work, Deep Research provides a summary of the results and a list of all sources used.

Users can set the research topic and modify the process not only through text but also by uploading files. However, OpenAI clarifies that Deep Research currently has limited analytical capabilities and may make mistakes (“hallucinate”).

Access and limitations

Due to the high resource demands of Deep Research, it is currently available with restrictions:

• ChatGPT Pro subscribers ($200 per month) — up to 100 requests per month.

• Plus and Team users — have “limited access.”

OpenAI plans to increase the limits in the future when a more efficient version of Deep Research becomes available.