Thibault Sottiaux, who leads OpenAI’s core product area including ChatGPT Work, announced on the night of August 25 that starting August 26, Plus accounts will once again be subject to a 5-hour usage limit across both ChatGPT Work and Codex. He had previously mentioned this plan, but it was postponed.
He cited two reasons:
- The 5-hour limit helps smooth out the load on OpenAI’s compute infrastructure, while in return keeping weekly usage allowances generous.
- Plus users are mostly newcomers or casual users; quite a few accidentally burn through their entire weekly limit and are left confused about what happened, resulting in a poor experience.
The most notable point lies in the final line: over the next few months, the 5-hour limit will not apply to the $100 Pro and $200 Pro tiers. This means the boundary between tiers is no longer just token count or speed, but the structure of usage time: Plus is restricted by time windows, while Pro remains unconstrained during this transition period.
Under the post, which garnered over 562,000 views, reactions quickly split in two directions. One questioned the logic: if Plus users are casual, why are they the ones driving compute pressure that needs to be controlled? The other was more blunt, from one user: “I’m not a casual user, I’m just poor,” which received hundreds of likes. Both pointed to the same conclusion: this policy is essentially re-tiering perks by price, rather than merely protecting the newcomer experience as framed in the announcement.
For those using Plus to run long-running agents via Work or Codex, the practical takeaway is clear: tasks running continuously for multiple hours will require rethinking execution pacing, or considering an upgrade before the new limit takes effect.