OpenAI Lifts the Five-Hour Cap on ChatGPT and Codex After a Weekend of Overloaded Demand
Plus and Business subscribers get a usage reset and a temporary reprieve from the rolling limit, as OpenAI tunes GPT-5.6 Sol to burn through allowances more slowly.
Key points
- OpenAI temporarily removed the five-hour usage limit on ChatGPT Plus and Business plans on Sunday after two days of intense demand.
- The company reset every paying subscriber's usage counter, giving them a clean slate.
- Product lead Tibo said GPT-5.6 Sol is being retuned to consume less of each user's allowance per task.
- Codex and ChatGPT share the same usage pool, which is what caused many users to hit the wall so quickly.
OpenAI has pulled back one of its most unpopular restrictions, at least for now.
On Sunday the company confirmed it is temporarily removing the rolling five-hour usage limit for paying subscribers on Plus, Business and Pro tiers. Everyone's current usage counter has also been reset to zero. BleepingComputer first reported the change.
A surge in demand over the previous 48 hours forced the move. Both ChatGPT and Codex, OpenAI's tool for writing and running code, draw from the same pool of allowed messages and tasks. When one side gets busy, the other runs out of room too. We've tracked Codex closely since our first report on the tool in June, including the credential-theft risk that comes with developer reliance on it.
"The last 48 hours of Codex and ChatGPT Work have been intense," OpenAI product lead Tibo wrote on X. "[We're] temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans."
What does the five-hour limit actually do?
It stops you working once you have sent too many messages in a short window.
ChatGPT counts everything you do inside a rolling five-hour period. Send too many prompts, run too many Codex jobs, and the service tells you to come back later. Weekly caps sit on top of that, varying by plan and model. For developers running Codex agents, which can chew through dozens of prompts on a single task, the five-hour ceiling has been a persistent complaint. Removing it, even temporarily, means a long coding session no longer ends because a timer expired.
GPT-5.6 Sol is being made lighter
OpenAI is also changing how much of your allowance each request eats.
"[We are] rolling out changes that will make GPT-5.6 Sol more efficient across the board and that will be reflected in less usage being used so that it can take you further," Tibo said. GPT-5.6 Sol is the current flagship model, the one most subscribers reach for on hard problems.
OpenAI hasn't spelled out the mechanics. The most likely explanation is lower token consumption per request: the model processes the same question using fewer of the small text chunks, called tokens, that the company uses to meter usage. Fewer tokens per task means more tasks before you hit the cap.
The one-time reset matters too. Anyone who had burned through their weekly allowance before Sunday now has fresh headroom, without waiting for the normal cycle to tick over.
Should you worry about the limits coming back?
Yes, because this fix is explicitly temporary.
Weekly caps still exist on plans that have them. The five-hour rule has been suspended and OpenAI has been clear it will return in some form if demand cools. Free-tier users are not mentioned in the announcement and appear unaffected by the reset. For paying customers the practical picture is straightforward: you have more room to work this week than you did on Friday, and each GPT-5.6 Sol prompt should cost less of that room. Put the long Codex session you were deferring on the calendar now.
The deeper issue is architectural. Sharing one usage pool across a coding agent and a chat interface was always going to create contention as both products grow. A temporary cap removal doesn't fix that. Watch whether OpenAI separates the pools or simply keeps adjusting the pressure valve when demand spikes.
