OpenAI has introduced a new AI agent named Deep Research, following the recent launch of Operator. This new tool is tailored for conducting extensive research through its chatbot, ChatGPT.
Utilising the OpenAI o3 model optimised for web browsing and Python analysis, Deep Research employs advanced reasoning to navigate and analyse text, images, and PDFs found online, as stated by OpenAI.
In their blog announcement, OpenAI highlighted that “Deep research is built for people who do intensive knowledge work in areas like finance, science, policy & engineering and need thorough & reliable research.” Furthermore, they noted its utility for consumers seeking personalised recommendations on purchases that necessitate careful evaluation.
To use Deep Research in ChatGPT, users simply select the “Deep Research” option from the message composer and input their queries.
Additionally, they can attach relevant files or spreadsheets for context. Upon submission, a sidebar will display the steps being undertaken along with the sources referenced during the research process.
OpenAI indicated that the time required for this AI agent to complete research can range from 5 to 30 minutes, depending on the complexity of the inquiry. Users will receive a notification once the research is finalised, allowing them to continue with other tasks in the meantime. The outcome will be presented as a comprehensive report within the chat interface.
According to OpenAI, while Deep Research is adept at delivering thorough and well-documented responses with verified sources, GPT-4 remains more suitable for real-time, multimodal discussions. For inquiries that demand depth and precision in specific domains, Deep Research provides an edge through its extensive exploration and citation capabilities.
OpenAI plans to incorporate features such as embedded images, data visualisations, and additional analytical tools to further improve report clarity and context.
Distinction from Operator
The primary distinction between Deep Research and Operator lies in their functionalities. Operator can handle a variety of repetitive tasks like filling out forms or ordering groceries, but it lacks the capability for in-depth research that Deep Research offers. The latter builds upon advanced reasoning skills to tackle complex problems encountered in professional settings and daily life.
The launch of this new feature arrives amid heightened competition from other AI models like DeepSeek, which has garnered attention for its advanced technology at a lower cost. OpenAI CEO Sam Altman acknowledged DeepSeek’s capabilities while asserting that OpenAI’s models would surpass them in quality.