OpenAI’s new “Deep Research” capability enhances the model’s ability to conduct multi-step research on the internet, allowing it to handle complex queries and synthesize information efficiently, as demonstrated by its performance on intricate tasks across various fields. The tool’s potential for real-world applications, such as aiding in medical decision-making, highlights its transformative impact on knowledge work and its role in advancing towards artificial general intelligence (AGI).
In a recent announcement, OpenAI introduced its latest capability called “Deep Research,” which significantly enhances the model’s ability to conduct multi-step research on the internet. This new feature allows users to input complex queries, and the model can take its time—ranging from five minutes to over half an hour—to search various sources and synthesize information. Deep Research excels in handling intricate questions across diverse subjects, as demonstrated by its performance on the “Humanity’s Last Exam,” where it achieved a notable accuracy of 26.6%, surpassing previous models.
The video highlights several expert-level tasks where Deep Research has proven to be exceptionally efficient. For instance, in a chemistry-related query, the model saved a researcher four hours of manual investigation by providing detailed insights into mixed gas sorption. Similarly, it assisted a linguist and a healthcare professional by saving them significant amounts of time in their respective fields. These examples illustrate the model’s capability to handle complex, domain-specific inquiries that typically require extensive knowledge and research.
A personal story shared by an OpenAI team member emphasized the transformative potential of Deep Research in real-world scenarios. After his wife was diagnosed with breast cancer, he utilized the tool to analyze her surgical pathology report and seek guidance on the necessity of radiation therapy. The model not only confirmed the oncologists’ opinions but also provided deeper insights and cited relevant studies, showcasing its ability to deliver personalized and accurate medical information. This experience underscored the model’s potential to empower individuals with critical health-related decisions.
The video also featured a demonstration of Deep Research in action, where team members posed various queries to the model. The tool was shown to ask clarifying questions before conducting its research, ensuring that it understood the user’s requirements. As it browsed the internet, Deep Research synthesized information from multiple sources, ultimately producing well-structured reports that included citations and formatted data. This capability is particularly beneficial for professionals who need to compile comprehensive analyses quickly.
OpenAI’s leadership expressed excitement about the future of Deep Research, emphasizing its role in the company’s roadmap toward artificial general intelligence (AGI). The ability of the model to think longer and autonomously while solving complex tasks is seen as a significant step forward. With plans to make Deep Research available across various tiers of service, OpenAI aims to democratize access to advanced research capabilities, potentially transforming how individuals and organizations approach knowledge work and decision-making in the future.