OpenAI today announced Deep Research, a new agent that uses o3 model's reasoning capabilities to synthesize large amounts of web information. ChatGPT users will be able to use the Deep Research feature to complete multi-step research tasks. ChatGPT Pro plan users can access this feature starting today with 100 queries per month. This new feature is coming to ChatGPT Plus and Team users in the near future, while ChatGPT Free tier users will have access to this feature later on.
In general, research analysts spend several hours to find, analyze, and synthesize hundreds of online sources to create a research report. ChatGPT Deep Research can now do the same task in just a few minutes. This new feature was trained on real-world tasks requiring browser and Python tool use. The final report from Deep Research will have clear citations and a summary of its thinking. OpenAI claims that this feature will be really effective at finding niche, non-intuitive information that would require browsing several websites.
Here's how ChatGPT users can use this new Deep Research feature:
- Go to the ChatGPT web experience. (Deep Research will be coming to mobile and desktop ChatGPT apps later this month.)
- In ChatGPT, after entering your query, select ‘Deep Research’ in the message composer.
- Attach files or spreadsheets to add context to their question.
- Once the work starts, a sidebar appears with a summary of the steps taken and sources used.
- Deep Research may take from 5 to 30 minutes to complete its work, depending on the query.
- Once the report is ready, a notification will be sent to the user.
- For now, the final output is available as a text report within the chat. But in the next few weeks, ChatGPT will also be adding embedded images, data visualizations, and other analytic outputs in these reports for additional clarity and context.
- In the future, ChatGPT will also have the ability for users to connect to more specialized data sources to make its output even more accurate and personalized.
Humanity’s Last Exam is a new evaluation that tests AI models across a broad range of subjects on expert-level questions. It has over 3,000 multiple-choice and short-answer questions across more than 100 subjects. OpenAI's Deep Research model scored a record 26.6% accuracy in this evaluation. OpenAI also claims that this model reached a new state-of-the-art score in GAIA, another public benchmark that evaluates AI on real-world questions.
Deep Research has the potential to significantly reduce the time spent on research tasks. Its ability to synthesize information and provide citations could be a valuable asset for many users. You can learn more about OpenAI's Deep Research feature here.
5 Comments - Add comment