Deep Research
Deep Research is an intelligent research tool released by OpenAI that autonomously plans and executes multi-step web information retrieval and analysis tasks. The tool conducts systematic investigations of complex questions, integrates information from multiple sources, and produces structured research reports. It is suited for scenarios requiring intensive information processing, including academic research, market analysis, and technical investigations.
Core Facts
Deep Research能够自主规划研究路径、在互联网上执行多轮搜索、阅读并综合数十乃至上百个网页内容，最终生成结构化的深度报告
The Deep Research feature decomposes a single user query into multi-step autonomous research tasks, invokes tools such as web search and document analysis, and uses large language model reasoning to synthesize and output a structured report.
Deep Research is an autonomous research agent built on the o3 model, launched in early 2025.
Timeline (last 90 days)
On platforms such as Coze and Dify, conditional branch nodes and loop nodes can be used to simulate the research loop process of Deep Research.
The high computational cost of deep reasoning capabilities constitutes a hidden barrier, making unlimited use by individual users economically impractical.
Compared to ordinary conversations, Deep Research mode consumes several to dozens of times more tokens, and a single task may take several minutes or longer to complete.
Deep Research is a collective term for a type of enhanced reasoning capability released by multiple AI companies in recent years, with OpenAI's identically named feature released in early 2025 being a representative example.
Research-type agents are the first to reach product-ready levels, because such tasks require processing large amounts of information, tolerate latency, benefit from search, and can be automatically verified through citations.
Perplexity's deep research mode (Pro Search / Deep Research) consumes far more quota per use than ordinary question-and-answer queries.
The zeroed-out endpoints do not mean the account quota is truly exhausted—the Deep Research feature still works and the user remains within Pro quota.
Gemini's Deep Research feature first presents a research plan for the user to confirm, then automatically browses dozens to hundreds of sources, and finally generates a multi-page report with citations, with the free tier allowing approximately ten or so uses per month.
Deep Research is an in-depth research feature launched by Google in the paid Gemini Advanced tier, with a Gemini Advanced subscription costing approximately $20/month.
The third-party channel interface demonstrated includes features such as file upload, creating Projects, creating Skills, Deep Research, and web search.
12 more timeline events
All Facts (20)
Deep Research能够自主规划研究路径、在互联网上执行多轮搜索、阅读并综合数十乃至上百个网页内容，最终生成结构化的深度报告
80%VerifiedThe Deep Research feature decomposes a single user query into multi-step autonomous research tasks, invokes tools such as web search and document analysis, and uses large language model reasoning to synthesize and output a structured report.
75%VerifiedDeep Research is an autonomous research agent built on the o3 model, launched in early 2025.
65%UnverifiedOpenAI、Google、Perplexity等公司都推出了各自的Deep Research功能
90%UnverifiedDeep Research是Gemini 2.5更新的重大功能，可以同时搜索和浏览数百个网站
85%UnverifiedOpenAI的Deep Research能够在5到30分钟内完成一项深度研究任务，自动浏览数十甚至上百个网页来源
80%UnverifiedDeep Research支持超过24种语言的播客输出
75%UnverifiedRAG的局限在于它本质上是单轮检索，无法处理需要多步推理、交叉验证的复杂研究任务
70%UnverifiedOpenAI launched the Deep Research feature in February 2025, as an exclusive feature for ChatGPT Pro subscribers (at $200 per month).
60%UnverifiedOn platforms such as Coze and Dify, conditional branch nodes and loop nodes can be used to simulate the research loop process of Deep Research.
50%UnverifiedCompared to ordinary conversations, Deep Research mode consumes several to dozens of times more tokens, and a single task may take several minutes or longer to complete.
50%UnverifiedThe high computational cost of deep reasoning capabilities constitutes a hidden barrier, making unlimited use by individual users economically impractical.
50%UnverifiedDeep Research is a collective term for a type of enhanced reasoning capability released by multiple AI companies in recent years, with OpenAI's identically named feature released in early 2025 being a representative example.
50%UnverifiedResearch-type agents are the first to reach product-ready levels, because such tasks require processing large amounts of information, tolerate latency, benefit from search, and can be automatically verified through citations.
50%UnverifiedPerplexity's deep research mode (Pro Search / Deep Research) consumes far more quota per use than ordinary question-and-answer queries.
50%UnverifiedDeep Research is an in-depth research feature launched by Google in the paid Gemini Advanced tier, with a Gemini Advanced subscription costing approximately $20/month.
50%UnverifiedGemini's Deep Research feature first presents a research plan for the user to confirm, then automatically browses dozens to hundreds of sources, and finally generates a multi-page report with citations, with the free tier allowing approximately ten or so uses per month.
50%UnverifiedThe effectiveness of multi-agent scaling is highly dependent on the task domain: math problems and in-depth research reports can be parallelized, while writing novels is nearly impossible to parallelize.
50%UnverifiedPlatforms such as Coze have encapsulated Deep Research capabilities in the form of workflow nodes.
50%UnverifiedThe core difference between Deep Research and single-pass retrieval Q&A lies in iterative reasoning, where each round of search results updates the agent's understanding of the problem and generates a more precise query for the next round.
50%