OpenAI Launches Deep Research: A New Agentic Capability for Complex Tasks

OpenAI has introduced a significant new feature called "deep research" in ChatGPT, designed to handle multi-step internet research tasks that would typically take humans many hours. This agentic capability allows users to provide a prompt, and ChatGPT will autonomously find, analyze, and synthesize hundreds of online sources to generate comprehensive reports at the level of a professional research analyst.

Technical Overview

• Model: Deep research is powered by a version of the upcoming OpenAI o3 model, which has been optimized for web browsing and data analysis.
• Capabilities:
  • Web Browsing: It can navigate and search the internet, including accessing visual content through a browser.
  • Data Analysis: The agent can interpret and analyze large volumes of text, images, and PDFs.
  • Reasoning: It uses advanced reasoning to pivot its approach based on the information it encounters during research.

Why Deep Research Matters

Deep research is particularly valuable for professionals in fields such as finance, science, policy, and engineering who engage in intensive knowledge work. The ability to synthesize existing knowledge is a crucial step toward creating new insights and innovations. This capability marks a significant milestone in OpenAI's broader goal of developing Artificial General Intelligence (AGI), which they envision as being capable of producing novel scientific research.

Key Updates and Enhancements

• April 24, 2025:

  • Increased query limits for different user tiers:
    • Plus, Team, Enterprise, and Edu users: 25 queries per month
    • Pro users: 250 queries per month
    • Free users: 5 queries per month
  • Introduction of a lightweight version of deep research powered by o4-mini, which is more cost-efficient while maintaining high quality. Once the limit for the full version is reached, queries will automatically switch to this lightweight version.

• February 25, 2025:

  • All Plus users gained access to deep research.

• February 5, 2025:

  • Deep research became available to Pro users in the United Kingdom, Switzerland, and the European Economic Area.

• July 17, 2025:

  • Enhanced capabilities with access to a visual browser as part of ChatGPT agent.
  • Users can select "agent mode" from the dropdown in the composer to access these updated features.

• February 10, 2026:

  • Connect deep research to any Management Control Program (MCP) or app.
  • Restrict web searches to trusted sites for authenticated, industry-standard sources.
  • Real-time progress tracking and the ability to interrupt with follow-up prompts or new sources.
  • Improved visual experience for easier start, track, and review of research from end to end.

How It Works

1. Initiation: Users provide a prompt in the ChatGPT interface.
2. Research: The agent conducts multi-step research, leveraging its optimized model to navigate and analyze web content.
3. Synthesis: It compiles findings into a comprehensive report, similar to what a human analyst would produce.
4. Review: Users can review the report, refine it with additional queries, or integrate it into their work.

Use Cases

• Finance: Conduct detailed market analysis and trend forecasting.
• Science: Synthesize research papers and data for literature reviews.
• Policy: Gather and analyze policy documents and regulatory frameworks.
• Engineering: Research technical specifications and best practices for project planning.

Conclusion

Deep research is a powerful tool that leverages advanced AI to streamline complex research tasks. By automating the process of finding, analyzing, and synthesizing information, it frees up valuable time for professionals to focus on higher-level tasks and innovation. As OpenAI continues to refine and expand its capabilities, deep research represents a significant step toward achieving AGI.