
OpenAI is introducing a new version of its ChatGPT deep research tool that it says is cheaper to operate. The “lightweight” model is available now to free users, as well as those on several of the company’s paid tiers.
Deep research (lightweight) is powered by a version of OpenAI’s o4-mini model, and the company says it’s “nearly as intelligent” as the original deep research mode, which produces lengthy reports including citations and summaries of its process. OpenAI adds that the lightweight version delivers responses that are “shorter while maintaining the depth and quality you’ve come to expect.”
While the standard deep research model remains exclusive to paid users, the lightweight version is rolling out to free users now, who can use it for five tasks per month. Paid users also get access, and it will kick in by default when they hit their rate limits on the original deep research, giving Team and Plus users a total of 25 monthly tasks across both versions, and Pro users 250. Enterprise and Education users will get access next week at the same limits as Team and Plus users.
Importantly for OpenAI, it says the new version is “significantly cheaper” to run, which matters all the more in a world where it spends “tens of millions of dollars” just to process users saying please and thank you.