Openai revealed Another modern aggressive function for chatgpt called Deep Research, which, he says, can act autonomously to “plan and perform a multi -stage trajectory to find the necessary data, retreat and respond to real -time information if necessary.”
Instead of simply generating the text, it shows a summary of its process on the sidebar, with quotes and a summary showing the process used for reference.
Users can ask questions using text, images and additional files, such as PDF files or spreadsheets to add context, and then it will take “from 5 to 30 minutes” to develop a response in the chat window, with promises, which in the future This IT will also be able to include built -in paintings and charts. Opeli also notes the restrictions on deep research, saying that maybe “hallucination” and invent facts, fight the difference between authoritative information and rumors, and register how certain should assess the answer.
Developing ways to make generative AI tools more useful and worth paying are future companies, such as OpenAI, promised agents, and claims that deep research is able to act at the level of research analyst. The demonstration film attached here begins with a request for changes in the retail industry over the past three years, with a response covering points and tables.
This function is strictly in line with the launch of the OPENAI operator, tools that can employ a web browser to perform tasks for you, and is similar to the Project Mariner Research prototype Google prototype in December. The Google tool is not yet available to the audience, but Deep Research launches “with the optimized version for Pro users”.
Opeli offers up to 100 questions per month to people paying a monthly fee of USD 200 and “limited access” for a plus, a team, and ultimately corporate users, calling the ability “very calculating”, requiring more application, the longer it requires that it requires a longer number of need. Examine something. He also says that all paid users will receive higher rate limits in the future, when a faster, more profitable version is available.
The press message says that the deep research power model has obtained a modern high level of accuracy on the AI comparative test, which was called “The last exam of humanity“, Which asks about answering questions at the level of experts. The OpenAI Deep Research model reached an accuracy of 26.6 percent thanks to the browsing and Python tools on, well above 3.3 percent of the GPT-4O, and the next highest shooter, his O3-Mini model (high) rated only on the text, at the level of 13 percent.