💻 OpenAI Has Introduced Deep Research—An Advanced Agentic System
Today, OpenAI revealed Deep Research, a novel agentic system designed to conduct extensive online research autonomously.
Key Features:
➡️ A single Deep Research prompt takes up to 30 minutes to process but delivers a full-fledged research paper on the topic with source references.
➡️ An OpenAI employee mentioned that in 10 minutes, the agent performs work that would take up to three hours for him.
➡️ It runs on a version of o3, optimized for analyzing various sources, and uses Python.
➡️ The agent scores 26.6% on today's most challenging benchmark, "Humanity's Last Exam." In contrast, the o3-mini presented at the end of last week scored only 13%.
➡️ Deep Research is already available with the Pro subscription ($200 per month) with a limit of 100 prompts per month. Soon, all paid subscribers will have access, but with a limit of 10 monthly queries.
Deep Research hallucinates much less than other OpenAI models, though it is not entirely free from this issue. According to OpenAI's lead researcher, Mark Chen, the model not only analyzes content but also creates it. "I think our ultimate aspiration is a model that can uncover and discover new knowledge by itself," he explains.
More on the topic:
