OpenAI has introduced Deep Research, a ChatGPT tool that automates complex multi-step internet research. The AI tool is designed to generate “expert-level” reports and is accessible to Pro users. With competition rising from Chinese startup DeepSeek and other domestic AI rivals, OpenAI continues to innovate by releasing new products. Deep Research operates independently to collect information from the internet efficiently, completing tasks in a fraction of the time it would take a human. Available for Pro users at a monthly cost of $200, the tool is targeted towards professionals in finance, science, policy, and engineering. Powered by a refined version of OpenAI’s o3 reasoning model, Deep Research autonomously scours the internet, gathers data from various sources, and generates fully cited reports. OpenAI’s chief research officer, Mark Chen, describes it as comparable to reports produced by analysts or experts in a field. Despite achieving a notable score of 26.6% on “Humanity’s Last Exam,” surpassing GPT-4’s score, OpenAI acknowledges that the model may still produce incorrect information and struggle to discern between fact and rumor. OpenAI encourages users to test Deep Research on challenging tasks that can be solved using internet resources. Additionally, OpenAI recently launched o3-mini, a cost-effective reasoning AI model, in response to the competitive landscape. These releases come shortly after OpenAI’s commitment to enhancing its models following the debut of DeepSeek’s R1 model, which posed a formidable challenge to OpenAI’s offerings. Last month, OpenAI unveiled its Operator AI agent, followed by the public launch of the text-to-video model Sora in December.
