Open AI releases agent that can do research

Open AI has released a Deep research feature based on the latest company’s reasoning model, the o3.

The new agent uses reasoning to synthesize information and complete multi-step research tasks.

In a blog post about the new release, Open AI claims that the agent can accomplish in “tens of minutes what would take a human many hours.”

For example, a user can type a prompt, such as how the retail industry has changed in the past three years, and ask to present information with bullets and tables as well as add their own data to the query.

After the prompt is given, ChatGPT analyzes internet sources and references the output to verify the information. Since the feature leverages reasoning, it may take from 5 to 30 minutes to complete the query.

The tool, which is available only for ChatGPT Plus, Team, and Pro subscribers, is aimed at people who work in areas like finance, science, policy, and engineering.

Users can access the feature by choosing ‘Deep research’ in the message composer.

‘Deep research’ is powered by a version of the latest o3 model, which Open AI previewed last December.

On Friday, the company announced the o3-mini, which is optimized for web browsing and data analysis. Like OpenAI’s o1-mini, the o3-mini can reason and excels at science, math, and coding. However, the latest model reportedly has reduced latency.

o3-mini will replace the o1-mini in the model picker, while Plus and Team users will have a limit of 50 messages per day with o1-mini to 150 messages per day with o3-mini.

Open AI has also released o3-mini’s API for developers. The model will support streaming, allowing developers to choose between three options – low, medium, and high – to optimize for their specific use cases.

The o3-mini doesn’t include visual reasoning, meaning that developers should use o1 for visual tasks.