 "OpenAI's Deep Research AI agent")
Microsoft-backed artificial intelligence start-up OpenAI has introduced a new AI agent within ChatGPT called "Deep Research." The company said that the new integrated tool is powered by a version of the company's upcoming o3 series reasoning model and enables users to conduct multi-step research on the internet for complex tasks. Additionally, OpenAI has launched its new "o3-mini" small reasoning model, which is also available for free-tier users of ChatGPT.
OpenAI Deep Research: Details
What is it?
Deep Research is OpenAI's new AI agent that can analyse and synthesise various online sources to generate a comprehensive report on a topic provided by the user. AI agents are essentially advanced systems that can perform multi-step tasks for users with minimal assistance. They are capable of gathering information from text, images, audio and more to autonomously achieve objectives based on predefined goals.
Also Read
Deep Research is not the first AI agent released by the company. OpenAI recently launched its Operator agent for performing web-based tasks such as booking flight tickets, ordering groceries and more. Google also offers a "Deep Research" tool for its AI chatbot Gemini, which can analyse information from the web on behalf of users.
How it works?
OpenAI's Deep Research agent is powered by a version of the upcoming OpenAI o3 model which the company said is optimised for web browsing and data analysis. The new agent utilises this model's reasoning capabilities to search, interpret and analyse text, images, and PDFs on the internet.
Users can select 'deep research' in the message composer of ChatGPT and enter their query to start with the research. They can also attach files or spreadsheets to add context to their question. Once the agent starts the process, a sidebar appears with a summary of the steps taken and sources used. OpenAI said that the entire process can take five to 30 minutes depending on the type of research. ALSO READ: Google reinforces AI with Gemini 2.0 model, deep research mode, and more
The final output is presented as a report within the chat. The company also said that soon these reports will have embedded images, data visualisations, and other analytic outputs for additional clarity and context.
What are its limitations?
OpenAI said that the Deep Research AI agent is still in its early stages and can occasionally hallucinate facts in responses or make incorrect inferences. As per the company's internal evaluation, the new AI agent makes these mistakes "at a notably lower rate than existing ChatGPT models," but does struggle with distinguishing authoritative information from rumours.
Availability
OpenAI said that it is rolling out an optimised version of the new Deep Research AI agent to Pro-tier subscribers with a usage limit of 100 queries per month. Additionally, the company is also working on bringing access to Plus and Team subscribers next, followed by Enterprise users.
OpenAI o3-mini model: Details
Besides the new AI agent, OpenAI has released a new small-scale model in its o3-series reasoning model. The latest o3-mini reasoning model is being integrated inside ChatGPT and its API services, while the company is making a version with rate limits available to free-tier users of ChatGPT. The company said that the new o3-mini model performs well in solving maths, science and coding problems, while maintaining low cost and reduced latency, similar to the company's o1-mini model. Additionally, o3-mini works with ChatGPT's Search feature to find up-to-date answers with links to relevant web sources.
The o3-mini model is now available for ChatGPT Plus, Team, and Pro subscribers, while Enterprise access is expected soon. OpenAI is also increasing the usage limit for Plus and Team users from 50 messages per day with o1-mini to 150 messages per day with o3-mini. As for free-tier ChatGPT users, an optimised version of the o3-mini model is available by selecting 'Reason' in the message composer.
