OpenAI has introduced a powerful new AI agent called “Deep Research.” The tool is designed to generate comprehensive reports by analyzing vast amounts of online data in minutes. OpenAI unveils a new AI tool called Deep Research to assist in synthesizing complex information. According to OpenAI, deep research can accomplish tasks in ten minutes, which would typically take several hours for a human researcher.
The AI startup’s announcement follows its recent decision to speed up product releases amid competition from China’s DeepSeek. Deep Research is powered by OpenAI’s advanced o3 model, a cutting-edge AI system optimized for reasoning and data synthesis.
Capabilities of Deep Research
OpenAI unveils a new AI tool called Deep Research, providing researchers with powerful tools for advanced research. Deep Research can search, analyze, and synthesize information from hundreds of online sources, including text, images, and PDFs. OpenAI claims that the AI tool can generate detailed research reports with citations for each claim. The company views this as a major step toward developing artificial general intelligence (AGI), an AI system that can perform intellectual tasks at a human level or beyond.
The tool is particularly beneficial for professionals in finance, science, and engineering. However, it can also be used for consumer research, such as evaluating products like cars and furniture.
Availability and Subscription Model
Deep Research is currently available only to users in the United States under OpenAI’s Pro subscription plan, which costs $200 per month. Subscribers can use the tool up to 100 times per month due to its high processing costs. The service is not yet accessible in the UK or Europe.
The AI agent functions within ChatGPT as a dedicated button. Once activated, Deep Research performs its analysis and provides a detailed report within five to thirty minutes, depending on the complexity of the task.
Competition with DeepSeek and AI Advancements
OpenAI unveils a new AI tool called Deep Research, thus setting new standards for AI-driven research efficiency. DeepSeek, a Chinese AI firm, has emerged as a key competitor, claiming that its R1 model rivals OpenAI’s o1 model while being more cost-efficient. This has fueled concerns about OpenAI’s position in the AI race. In response, OpenAI CEO Sam Altman assured that the company will deliver superior models and accelerate product launches.
Deep Research is OpenAI’s second AI agent, which was launched within a month. Previously, the company introduced Operator, an AI tool designed to handle web browsing tasks, such as booking reservations and online shopping.
Development and Performance
Deep Research is built on OpenAI’s upcoming o3 model, optimized for browsing and data analysis. The AI was trained using real-world tasks across multiple domains that require complex reasoning and information retrieval. The company utilized reinforcement learning to improve the tool’s ability to navigate the web, backtrack when needed, and refine its search results dynamically.
Benchmark tests have highlighted Deep Research’s capabilities. The tool achieved a record accuracy of 26.6% on “Humanity’s Last Exam,” a benchmark that assesses AI performance across expert-level topics. Additionally, the AI model excelled in the GAIA test, which evaluates AI reasoning, multi-modal fluency, and web browsing skills.
Challenges and Future Enhancements
Despite its strengths, OpenAI acknowledges that Deep Research has limitations. The AI may struggle to differentiate authoritative sources from unreliable information and lacks confidence calibration, meaning it does not always communicate uncertainty accurately.
OpenAI plans to enhance Deep Research by embedding images, data visualizations, and other analytical elements directly into reports. Users will also be able to integrate specialized data sources for more precise and personalized outputs.
Deep Research represents a significant advancement in AI-driven research, but concerns remain about reliance on AI-generated content without human verification. Experts caution that while the tool is powerful, thorough fact-checking remains essential.
