OpenAI has introduced a lighter version of Deep Research, the advanced feature that scours the web recursively, reassesses information, and takes between five to 30 minutes to deliver a response. Originally designed to boost performance on Humanity’s Last Exam AI benchmark, Deep Research is now available in a more accessible form for all users — including those on the free plan.

The full Deep Research tool runs on OpenAI’s upcoming o3 model, while the new lightweight version is powered by o4-mini. According to a post on X (formerly Twitter), OpenAI claims the o4-mini model is „significantly cheaper to serve“ while maintaining accuracy levels relatively close to the full version. The lightweight model achieves a 45.6% score on OpenAI’s BrowseComp benchmark, compared to 51.5% for the standard tool available to paid users.

>>>HB522628EFW for Honor GS3 MUS-B19 1ICP6/26/27

For Pro users, OpenAI has expanded access by offering an additional 125 Deep Research queries through the lightweight model once their initial 125 queries are exhausted. Meanwhile, Plus and Team users will see their monthly quota increase from 10 tasks to 25. OpenAI also announced that Enterprise and Education users will receive an extra 15 lightweight queries starting next week.

Free users can now try Deep Research Lite as well, although they are limited to five tasks per month. While this limited access may be sufficient for many of ChatGPT’s estimated 400 million users, those who rely heavily on deep research capabilities will likely need to upgrade to a paid plan, which ranges from $20 to $200, to unlock more usage.