In a recent announcement from OpenAI, a new feature called Deep Research Mode has been introduced, promising to significantly enhance the way we approach complex research tasks. This advanced tool is set to support knowledge work by providing comprehensive, fully cited research papers that can be tailored to a wide range of applications. Let’s find out more about OpenAI’s Deep Research!
Table of contents
What is Deep Research Mode?
Deep Research Mode is an AI model designed to conduct multi-step research on the internet. Unlike traditional models that provide quick answers, Deep Research Mode takes a more thorough approach. It can spend up to 30 minutes or more on a single query, allowing it to discover, synthesize, and reason through vast amounts of information. This extended research time enables the model to produce detailed, well-cited reports that are as good as those produced by human analysts.
Key Features of Deep Research Mode
- Comprehensive Research: Deep Research Mode goes beyond simple searches. It opens web pages, reads through content, and even analyzes images, tables, and PDFs to gather information. This multi-step process ensures that no stone is left unturned.
- Adaptive Reasoning: The model adapts its research plan as it uncovers new information. This dynamic approach allows it to refine its search parameters and focus on the most relevant data.
- Long-Form Output: Deep Research Mode produces comprehensive reports that include detailed analysis, formatted tables, and clear recommendations. Users can specify the format and structure of the output, making it highly customizable for various use cases.
- Latency-Free: By removing latency constraints, Deep Research Mode can take the time needed to produce high-quality results. This is a significant departure from traditional models that prioritize speed over depth.
Also Read: 5 Ways to Use ChatGPT’s Scheduled Task Feature
How it Works?
Deep Research Mode is powered by a fine-tuned version of OpenAI’s upcoming o3 reasoning model. It uses end-to-end reinforcement learning on hard browsing and reasoning tasks, enabling it to plan and execute multi-step trajectories. The model can browse user-uploaded files, perform calculations, create images and plots, and even embed these elements in its final response. It cites specific sentences and passages from its sources, ensuring that the information is both accurate and verifiable.
Humanity’s Last Exam
Deep Research performed exceptionally on Humanity’s Last Exam, a benchmark from the Center for AI Safety and Scale AI, which tests models’ capabilities across a wide range of expert subjects. It achieved an impressive 26.6% accuracy, setting a new high score and demonstrating its advanced reasoning and research abilities.
| Model | Accuracy (%) |
|---|---|
| GPT-4o | 3.3 |
| Grok-2 | 3.8 |
| Claude 3.5 Sonnet | 4.3 |
| Gemini Thinking | 6.2 |
| OpenAI o1 | 9.1 |
| DeepSeek-R1* | 9.4 |
| OpenAI o3-mini (medium)* | 10.5 |
| OpenAI o3-mini (high)* | 13.0 |
| OpenAI deep research** | 26.6 |
* Model is not multi-modal, evaluated on text-only subset.
**with browsing + python tools
GAIA
Deep Research achieved a top performance on the test, reaching an accuracy of 26.6%. This result placed it significantly ahead of other models, showcasing its superior capability in reasoning and complex task execution
Expert Level Tasks
In an internal evaluation of expert-level tasks across a range of areas, deep research was rated by domain experts to have automated multiple hours of difficult, manual investigation.
Also Read: OpenAI’s Operator – ChatGPT Like Moment for AI Agents
How to Access Deep Research Mode?
Access to Deep Research Mode in ChatGPT is currently optimized for Pro users, with up to 100 queries per month. This version is compute-intensive, as the longer it takes to research a query, the more inference compute is required.
Also Read: OpenAI o3-mini: Performance, How to Access, and More
Applications of Deep Research Mode
Deep Research Mode has a wide range of applications across different domains:
Product Research
Prompt: “Help me find iOS and android adoption rates, % who want to learn another language, and change in mobile penetration, over the past 5 years, for top 10 developed and top 10 developing countries by GDP. Lay this info out in a formatted report, a table on metrics, and include recommendations on markets to target for a new translation app from ChatGPT, focusing on markets ChatGPT could better expand into“
Prompt: “Penetration as a percentage, and look at overall usage. Make your best assumptions on the rest!”
Purchase Research
Prompt: “I want to buy some skis, for skiing in Japan. Format this as a report with a nice table at the end.”
Knowledge Work
For enterprises and professionals, Deep Research Mode can streamline processes and enhance productivity. It can be used for market research, academic studies, and even day-to-day tasks that require extensive web browsing. For example, a product manager can use it to gather data on market trends, consumer preferences, and competitive analysis, all within a matter of minutes.
Consumer Use
Deep Research Mode is not just for professionals; it can also be incredibly useful for consumers. Imagine needing to make a significant purchase, like buying a new set of skis. Deep Research Mode can scour the internet for reviews, specifications, and user feedback, providing a comprehensive report that helps you make an informed decision.
Academic Research
In academic settings, Deep Research Mode can assist researchers in finding and synthesizing information from various sources. It can help in identifying relevant papers, analyzing data, and even suggesting new research directions. This can significantly reduce the time and effort required for literature reviews and data analysis.
You can find some of these applications here.
Conclusion
Deep Research Mode represents a significant advancement in AI-driven research. It will be interesting to see how people utilize Deep Research Mode once it becomes available to Plus users in the coming weeks. This feature could potentially streamline complex research tasks and provide valuable insights across various domains.
Stay updated with the latest happening of the AI world with Analytics Vidhya News!
Hello, I am Nitika, a tech-savvy Content Creator and Marketer. Creativity and learning new things come naturally to me. I have expertise in creating result-driven content strategies. I am well versed in SEO Management, Keyword Operations, Web Content Writing, Communication, Content Strategy, Editing, and Writing.
Free Courses
Nice article
Thanks!