an intelligent agent (agent) that helps users efficiently process massive amounts of online information. This function has already been made available to Pro users today and will subsequently be rolled out to Plus and Team users. (Feels like the Pro account was worth buying..)
What is Deep Research?
to complete.
Its core capabilities include:
：Formulate search strategies and plan research paths like human researchers. ：Automatically retrieve, analyze, and summarize massive amounts of data on the Internet to generate high-quality research reports. ：During the research process, Deep Research adjusts its strategy based on real-time information to ensure the accuracy and completeness of the conclusions. : It can parse files uploaded by users, perform data visualization using Python, and embed charts and images from webpages.
Internet user case studies
Internet user Dan Shipper @danshipper
The tests conducted using Deep Research include:
(from 2020 to present) Analyze Tolstoy's character descriptions and derive his perspective on human nature (Corporate Annual Financial Report), to uncover potential unreported financial anomalies 👕 Starting from a few photos, research and recommend an entire new wardrobe pairing
Current limitations
If the task goes off track, there is currently no option to interrupt it midway,
Netizen Ethan Mollick (@emollick)
who actively looks for clues and works around research obstacles."
. This marks
Official case examples
Business | Commerce
Help me find iOS and Android adoption rates, % who want to learn another language, and change in mobile penetration, over the past 10 years, for top 10 developed and top 10 developing countries by GDP. Lay this info out in a table and separate stats into columns, and include recommendations on markets to target for a new iOS translation app from ChatGPT, focusing on markets ChatGPT is currently active in.
Please help me find the adoption rates of iOS and Android among the top 10 developed and developing countries by global GDP over the past 10 years, the percentage of users who wish to learn another language, and changes in mobile device penetration rates. Please organize this information into a table, listing each data item in separate columns, and provide market positioning recommendations for new iOS translation apps targeting ChatGPT's currently active markets.
Needle in a Haystack | Needle in the haystack
There is a TV show that I watched a while ago. I forgot the name but I do remember what happened in one of the episodes. Can you help me find the name? Here is what I remember in one of the episodes:
Two men play poker. One folds after another tells him to bet. The one who folded actually had a good hand and fell for the bluff. On the second hand, the same man folds again, but this time with a bad hand. A man gets locked in the room, and then his daughter knocks on the door. Two men go to a butcher shop, and one man brings a gift of vodka. Please browse the web deeply to find the TV show episode where this happened exactly.
I once watched a TV series, but I forgot its name. I only remember the plot of one episode. Can you help me find the name of this series? Here's what I remember:
Two men were playing poker. One chose to fold after the other asked him to bet. In fact, he had a good hand, but fell for the other's bluff. In the second round, he folded again, but this time his hand was indeed bad. A man was locked in a room, and then his daughter knocked on the door. Two men went to a butcher shop, and one of them brought a bottle of vodka as a gift. Please conduct an in-depth search online to find the TV episodes that accurately match these plotlines.
Medical Research | 医学研究
Do a deep dive into attempts to improve the reprogramming efficiency of OSKM by directly modifying the protein sequences of the four Yamanaka factors. List all relevant papers you find, the authors, the methods used, and the results. Study the patterns in the changes to the proteins and corresponding results across the papers and list the top 3 domains that scientists modify to increase efficiency, and why they believe these changes are effective.
, as well as their reasoning for why these modifications were effective.
UX Design | User Experience Design
Find evidence that shows that buttons with icons & labels are more usable than buttons without labels, or labels without icons. I know there’s been a lot of user studies on it, would love to see a detailed report along with a high-level, once definitive answer on the effectiveness.
demonstrates the effectiveness of this design.
Shopping | 购物
I’m looking for the perfect snowboard. I will be riding primarily in Hokkaido around twice a month during the winter season. I enjoy groomed runs but also want a board that can handle some fresh powder on occasion. I prefer a versatile all-mountain or freestyle board with a medium flex, something that’s stable for carving yet maneuverable in variable conditions. I want something with a fresh, citrus color palette that will pop on the slopes. My budget is mid-range to slightly premium, and I’d like suggestions on specific brands and models that are accessible in Japan. Please explain why each recommended board suits my requirements. Also, include any tips or considerations for riding in Hokkaido’s unique snow conditions. Include images of the items and format it in an easy-to-read table.
which not only maintains stability during carving but also ensures good control in various snow conditions.
Format for organizing information.
General Knowledge | Common Sense
What’s the average retirement age for NFL kickers?
What is the average retirement age for NFL kickers?
How does Deep Research work?
across multiple professional domains.
During the training process, it learned:
: to develop the optimal search strategy based on query content and continuously adjust the direction. : when encountering contradictory or missing information, automatically backtrack and adjust the research method. : It can directly analyze uploaded files, draw charts, generate visualized data, and embed the images or information sources into websites.
set new records.
Deep Research's Breakthrough in AI Evaluation
, it has achieved
1️⃣ Humanity's Last Exam
, setting a new high for this test.
🔍 What is Humanity’s Last Exam?
, evaluating the performance of AI on professional-level questions. , comprehensively testing AI's professional knowledge capabilities in various fields.
📊 Deep Research's performance
have seen the most significant progress. , with the ability to actively search for professional information rather than relying solely on existing training data.
2️⃣ GAIA Evaluation
(an open evaluation that measures AI's ability to solve real-world problems), Deep Research
🔍 What is the GAIA Evaluation?
standard test. The difficulty of the questions is divided into three levels, testing AI's
📊 Breakthrough in Deep Research
It has demonstrated strong adaptability across all difficulty levels of the GAIA evaluation, surpassing all previous AI models. Deep Research outperforms traditional AIs in handling complex problems.
Deep Research's performance on expert-level tasks
especially excelling in the following areas:
Chemistry（Chemistry） 
Linguistics（Linguistics） 
Healthcare（Healthcare） 
which can significantly improve efficiency and reduce the burden of manual research.
Limitations of Deep Research
Although Deep Research has unlocked many new capabilities, it is still in its early stages and has certain
(Hallucination): Although its error inference rate is significantly reduced compared to the existing ChatGPT model, it may still generate inaccurate facts. and There are still shortcomings in this aspect, and it may not be able to accurately express the uncertainty of research conclusions. : There may be slight errors in reporting and citation formats. OpenAI anticipates that these issues will be optimized with increased usage over time.
Deep Research access permissions
Currently, OpenAI has adopted a phased rollout strategy:
can use it, with a maximum of per month will gain access in the coming weeks. is expected to be launched in the coming months. are still waiting for support; OpenAI is optimizing compliance and technical infrastructure.
The future development of Deep Research
. In addition, OpenAI plans:
，Offer more professional and personalized research support. to automatically complete more complex tasks.
assist users