"The last 48 hours of Codex and ChatGPT Work have been intense," OpenAI product lead Tibo wrote on X, announcing a temporary change to how the company measures usage of its most powerful model.
OpenAI removes the five‑hour cap for Plus, Pro and Business plans
On Sunday, OpenAI confirmed it is temporarily lifting the five‑hour usage restriction that applied to its Plus, Pro and Business plans. The company also issued a one‑time reset of current usage for all customers. According to Tibo's post, the decision came after demand for the company's flagship model surged over the preceding 48 hours.
That five‑hour window had been part of ChatGPT's rolling usage billing—users could be forced to stop when they exhausted that short window even if they had remaining quota on weekly limits. With the shorter restriction removed, customers "are no longer forced to stop working simply because you have exhausted that five‑hour window," the company said in its explanation of the change.
How usage was counted: Codex, ChatGPT Work and shared limits
OpenAI noted that Codex and ChatGPT count local messages and cloud‑based tasks against a shared usage limit. In practice this meant coding activity and other cloud tasks drew from the same pool of available usage. ChatGPT's usual rolling five‑hour window remained a gating mechanism for many users, alongside plan‑dependent weekly caps for different models.
The temporary removal of the shorter restriction affects all Plus, Pro and Business subscribers, altering that immediate gating mechanism but not necessarily changing longer‑term weekly limits tied to specific plans or models.
GPT‑5.6 Sol: the efficiency changes that followed the surge
Alongside the usage reset and the removal of the five‑hour cap, OpenAI said it is "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 wrote. The company did not provide technical detail in that post.
The available explanation in the announcement suggests the efficiency gains likely reduce token consumption, though OpenAI did not confirm the mechanism. The company framed the change as a way to stretch users' available usage rather than as an elimination of limits: the one‑time reset "gives you significantly more room to use its flagship model for coding and agentic work, although it does not necessarily make GPT‑5.6 Sol completely unlimited."
What this means for Codex and ChatGPT Work users, Plus/Pro/Business subscribers, and enterprises using agentic workflows
- Codex and ChatGPT Work users: Developers doing coding work will see immediate relief from interruptions tied to the five‑hour rolling cap; shared usage between local messages and cloud tasks will still draw from the same pool, but the reset and efficiency changes provide extra headroom.
- Plus, Pro and Business subscribers: These plan holders are the direct beneficiaries of the temporary removal and the one‑time reset. They gain more continuous access to GPT‑5.6 Sol in the short term, but they retain exposure to any remaining weekly limits the plans impose.
- Enterprises running agentic workflows: OpenAI specifically noted the extra room helps for "agentic work." Teams orchestrating multi‑step or automated agent tasks will likely be able to run more iterations before hitting usage constraints, while still depending on the company’s ongoing adjustments to efficiency.
Short horizon, open questions
OpenAI framed the measures—temporary removal of the five‑hour cap, a one‑time usage reset, and an efficiency rollout—as immediate responses to a sharp surge in demand. They are explicitly temporary adjustments rather than a permanent removal of usage controls: the company cautioned the changes "do not necessarily" make GPT‑5.6 Sol unlimited.
What remains unsettled in the company’s public remarks is the precise technical path behind the efficiency gains. OpenAI indicated the reductions in reported usage likely stem from lower token consumption but did not provide confirmation or metrics. For organizations planning heavy coding or agentic workloads, the near‑term effect is clearer—fewer interruptions and more available quota—while the longer‑term calculus will depend on how OpenAI balances system capacity, demand and policy.
