Openi Is revealed Another new agent feature for a chat called Deep Research, which says it can work autonomously to work and execute a multi-step trajectory, which to find the required data, real time Where necessary to react and react to the information of. “
Instead of generating text only, it shows the summary of its process in a sidebar, showing the process used for quotes and references.
Users can ask questions using lessons, images and additional files like PDFS or spreadsheet to add reference, and then to develop the response provided in the chat window, “will take” anywhere from 5 to 30 minutes “, in which the future It will be with the promise that will also be able to include embedded images and charts. Openai also notes boundaries for deep research, stating that it can “sometimes habitti” and create facts, struggle with the difference between official information and rumors, and it can Can register how it should be a reaction.
More useful for generic AI tools and developing methods for paying are future companies like OpenIAI Promised agentsAnd it claims that intensive research is capable of operating at the level of a research analyst. The demo video involved here begins with a request for information about the change in retail industry in the last three years, consisting of a reaction that includes bullet points and tables.
This feature follows Openai closely Inauguration of operatorA tool that can use a web browser to complete tasks for you, and the project marinner research shown in December is similar to Google. Google's tool is not yet available to the public, but launching deep research “today with a version adapted to Pro users.”
Openai is offering 100 questions per month for those paying for those paying $ 200 monthly fee and finally, promised to “limited access”, “very calculated intensive” ability to call, team, and finally, promised “limited access” For, more time is required. To do some research. It also states that all paid users will get high rate limitations in future when a sharp, more cost -effective version is available.
A press release stated that model powering deep research scored a new high for accuracy on an AI benchmark, called “called”Last exam of humanity“One who asks for answers to expert level questions. Openai deep research model reached 26.6 percent accuracy with browsing and python tools, above 3.3 percent of GPT-4o, and the next highest scorer, its O3-MINI (High) The model evaluated only on the text, at 13 percent.