OpenAI unveils a new ChatGPT agent for ‘deep research’

OpenAI launches ChatGPT agent for complex research with limitations.

: OpenAI introduces 'deep research,' an AI agent in ChatGPT for detailed studies in finance, science, and more. Available to ChatGPT Pro users with 100 queries monthly, it targets thorough analyses over quick answers. While initially text-only, future updates include images and specialized data sources. Using the o3 reasoning model, deep research showed an accuracy of 26.6% in Humanity’s Last Exam, exceeding other models but with noted limitations.

OpenAI has unveiled a new AI agent called 'deep research' within its ChatGPT platform, specially designed for intricate research in fields like finance, science, policy, and engineering. Initially available to ChatGPT Pro users with a limit of 100 queries per month, this tool is meant for instances requiring thorough information synthesis from diverse sources.

At present, deep research interacts only via text, but plans are in place to incorporate multimedia features such as images and data visualizations. The o3 reasoning AI model is the backbone of this tool, harnessed through reinforcement learning and optimized for web browsing and data analysis.

Tested with Humanity’s Last Exam, the model achieved 26.6% accuracy, indicating room for improvement albeit outperforming others like Gemini Thinking at 6.2%. Despite its promise, it struggles with distinguishing reliable information and reporting uncertainties, yet its well-documented outputs offer value for users needing rigorously cited data.