OpenAI has unveiled Deep Research, a new ChatGPT agent designed to support users who require thorough, precise, and multi-source research. Rather than providing quick summaries, Deep Research assists with complex investigations across finance, science, policy, and engineering fields. It provides detailed insights to help consumers make significant purchases, such as cars or home appliances.
Available now for ChatGPT Pro users (with a 100-query limit per month), the feature will soon expand to Plus and Team users, followed by Enterprise. However, its release is limited geographically, with no current timeline for users in the U.K., Switzerland, or the EEA.
Deep Research is currently a web-exclusive tool, but mobile and desktop app support is expected later this month. When submitting a query, users can attach files or spreadsheets, and responses may take 5 to 30 minutes to generate. OpenAI stated that the feature will soon include data visualisations, embedded images, and connections to specialized data sources. In addition, it will include subscription-based and internal resources.
Deep Research is powered by a custom-trained version of OpenAI’s o3 model to improve accuracy, optimised for web browsing and data analysis. This model was trained through reinforcement learning, enabling it to interpret better, analyse, and synthesise information from various sources, including user-uploaded files. It can also generate and embed graphs, reference sources with specific citations, and summarise its reasoning.
Despite its improvements, OpenAI acknowledges that Deep Research isn’t flawless. It may misinterpret information, struggle with credibility assessment, or present uncertainty poorly. Additionally, formatting and citation accuracy may sometimes fall short.
With this launch, OpenAI competes with Google’s AI-powered “Deep Research” tool, which was announced just months earlier.