OpenAI brings Deep Research to ChatGPT's free users
ChatGPT's latest feature recursively scours the web, reassesses results, and requires anywhere from five to 30mins to compute a response. Deep Research, the latest ruse from OpenAI to score higher on Humanity's Last Exam AI benchmark test, now exists in a lighter version available to all users including those on the free plan.
The original deep research tool is powered by OpenAI's upcoming o3 model. In contrast, the lightweight version runs on OpenAI o4-mini. In a post on X, formerly Twitter, the company asserts that this model is "significantly cheaper to serve" despite being somewhat close in accuracy to the standard deep research tool, which is restricted to ChatGPT users on the paid tiers.
Pro users now get an additional 125 deep research queries with the lighter model when they exhaust their initial 125. For Plus and Team users, the quota is increasing from 10 tasks per month to 25. OpenAI says the same additional 15 queries with the lightweight model will be coming to Enterprise and Edu users in the following week.
Other users who rely on the free plan can also access the lightweight version of deep research, although they have to limit their queries to five tasks every month. While this may be adequate for the vast majority of ChatGPT's 400 million users, nearly anyone who actually finds deep research useful may have to shell out between $20 - $200 to increase their usage.
Source(s)
