OpenAI has released a new AI agent called “Deep Research”, which allows users to perform complex research tasks using ChatGPT.
OpenAI announced in a blog post that all ChatGPT Pro subscribers will receive access to the new feature today, with a limit of 100 monthly queries. The announcement added that “Plus and Team users will get access next, followed by Enterprise.”
According to the performance benchmarks released by the team, Deep Research achieved an accuracy of 26.6% on Humanity’s Last Exam. Hence, it is superior to OpenAI’s previous AI models for tasks involving high-level knowledge.
Each response process takes 5 to 30 minutes, and users can include supporting files or spreadsheets in their search queries. Currently, only ChatGpt web supports the Deep Research feature, and it will become available on mobile and desktop apps later this month.
Furthermore, the team has clarified that every Deep research response will include references and an explanation of the process used to derive conclusions from research queries. The goal is to increase users’ trust in the results and allow effortless verification of data.
As of now, deep research’s functionality is limited to text-based results. In future updates, OpenAI intends to incorporate image and data visualization systems along with analytical tools extending its “access to subscription-based or internal resources”.
