RT team
OpenAI has introduced a fresh ChatGPT agent, dubbed “deep research,” designed to support users in performing extensive, detail-oriented investigations using its renowned AI-powered chatbot. Tailored for professionals in fields such as finance, science, policy, and engineering, this new tool delivers comprehensive, accurate, and dependable research for complex tasks. Even consumers seeking well-informed decisions on purchases like vehicles, appliances, or furniture can benefit from its capabilities.
Unlike typical quick-answer responses, ChatGPT deep research aggregates data from diverse websites and resources, ensuring thorough analysis. Initially available to ChatGPT Pro subscribers with a limit of 100 queries per month, the feature will soon expand to include Plus and Team users, followed later by Enterprise access. An anticipated rollout in about a month promises higher query allowances, although users in the U.K., Switzerland, and the EEA will experience a delayed launch.
Using this tool is simple: select “deep research” in the chat composer, enter your query, and optionally attach files or spreadsheets. Results are delivered as text only after a processing time of five to thirty minutes, with mobile and desktop integrations arriving shortly. Future updates will add images, data visualizations, and connections to specialized data sources.
To enhance reliability and minimize common AI errors, OpenAI employs a specialized version of its o3 reasoning model, trained via reinforcement learning on real-world tasks. Early testing on Humanity’s Last Exam shows a 26.6% accuracy, outperforming similar models, despite occasional mistakes and formatting issues. This innovative tool marks a significant step forward in AI research capabilities remarkably.
