OpenAI Rolls Out Deep Research for all ChatGPT Plus Subscribers

OpenAI's Deep Research: A Game-Changer for AI Research

OpenAI has unleashed Deep Research for all ChatGPT Plus subscribers, promising a significant boost in research efficiency. Having tested similar features from competitors like Gemini, Grok 3, and Perplexity, I can confidently declare OpenAI's Deep Research as the superior option. This blog delves into its capabilities.

Table of Contents

• What is Deep Research Mode?
• Key Features of Deep Research Mode
• How Deep Research Works
  • Humanity's Last Exam Benchmark
  • GAIA Performance
  • Expert-Level Task Automation
• Accessing Deep Research Mode
• Deep Research in Action
  • Task 1: Website SEO Analysis
  • Task 2: Generative AI Salary Report
• Applications of Deep Research Mode
  • Product Research
  • Purchase Research
  • Professional/Enterprise Use
  • Consumer Applications
  • Academic Research
• Conclusion

What is Deep Research Mode?

Deep Research is an advanced AI model designed for in-depth internet research. Unlike models offering quick answers, it meticulously investigates queries, spending up to 30 minutes or more per request. This allows it to synthesize vast amounts of information, generating detailed, well-cited reports rivaling those of human analysts.

Key Features of Deep Research Mode

• Comprehensive Research: Deep Research goes beyond simple keyword searches. It actively browses web pages, analyzes content, and even interprets images, tables, and PDFs.
• Adaptive Reasoning: Its research strategy dynamically adapts based on discovered information, refining search parameters for optimal results.
• Detailed Output: It creates thorough reports with detailed analysis, formatted tables, and clear recommendations, offering customizable output formats.
• Unconstrained Latency: Prioritizing quality over speed, Deep Research dedicates the necessary time to deliver high-quality results.

How Deep Research Works

Deep Research leverages a refined version of OpenAI's upcoming o3 reasoning model. Utilizing end-to-end reinforcement learning on challenging browsing and reasoning tasks, it plans and executes multi-step research processes. It can process user-uploaded files, perform calculations, generate images and charts, and seamlessly integrate these elements into its final response. Crucially, it provides citations to sources, ensuring accuracy and verifiability.

Humanity's Last Exam Benchmark

Deep Research excelled on Humanity's Last Exam, a benchmark from the Center for AI Safety and Scale AI assessing expertise across various domains. Its 26.6% accuracy set a new record, highlighting its advanced reasoning and research capabilities.

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 Performance and Expert-Level Tasks

Deep Research achieved top performance on GAIA, further demonstrating its superior reasoning and complex task execution capabilities. Internal evaluations also showed it automated hours of difficult manual research across various expert fields.

Accessing Deep Research Mode

Deep Research is available to ChatGPT Plus and Pro subscribers. Simply open a new chat, locate the "Deep research" icon, and begin.

Deep Research in Action

Here are examples showcasing Deep Research's capabilities:

Task 1: Website SEO Analysis

(Prompt and Output shown as a video example in the original text. The summary below is provided.)

Deep Research provided a comprehensive SEO analysis, offering tailored strategies for a specific blog, addressing challenges like AI content saturation, Google dependency, and ranking issues. The response was well-structured, evidence-based, and offered actionable advice.

Task 2: Generative AI Salary Report

(Prompt and Output shown as a video example in the original text. The summary below is provided.)

The report provided detailed salary comparisons across various countries, experience levels, skills, and industries, including cost-of-living adjustments and visualizations.

Applications of Deep Research Mode

Deep Research's versatility extends across numerous fields:

• Product Research: Market trend analysis, consumer preference research, competitive analysis.
• Purchase Research: In-depth product comparisons, reviews, and recommendations.
• Professional/Enterprise Use: Streamlining research for market analysis, academic studies, and various knowledge work tasks.
• Consumer Applications: Facilitating informed decision-making for significant purchases.
• Academic Research: Assisting with literature reviews, data analysis, and identifying research directions.

(Examples of specific prompts and outputs for Product and Purchase Research are shown as images in the original text. Those images are not reproduced here.)

Conclusion

Deep Research signifies a major leap forward in AI-powered research. Its potential to streamline complex tasks and provide valuable insights across various domains is substantial. It will be fascinating to observe how users leverage this powerful tool.

The above is the detailed content of OpenAI Rolls Out Deep Research for all ChatGPT Plus Subscribers. For more information, please follow other related articles on the PHP Chinese website!