If your team uses ChatGPT Plus for coding or autonomous task work, something changed this week. On August 25, 2026, OpenAI quietly restored a rolling 5-hour cap on both Codex and ChatGPT Work for Plus subscribers, reversing a brief period in which only the weekly quota applied.

Thibault Sottiaux, the technical lead for Codex and ChatGPT Work at OpenAI, announced the change on X on August 24, giving users a day’s notice before the limit kicked back in.

What Changed and Why

The cap is not a new concept. OpenAI had originally removed it weeks earlier as an experiment, but the outcome was not what the company hoped for. According to Sottiaux, users on the Plus plan tend to be more casual or newer to the tools. Without the rolling cap, many would burn through their entire weekly quota in a single session and then wonder why the tools were unresponsive for the rest of the week.

The 5-hour rolling window works as a consumption ceiling that resets progressively, sitting alongside the existing weekly usage quota. Hit the window limit and you need to wait for it to roll off before resuming heavy usage.

The rationale is operational. OpenAI framed it as load smoothing at the compute level, keeping the weekly allocation generous while preventing spikes that degrade the experience for everyone on the tier.

Who Is and Is Not Affected

The change only applies to ChatGPT Plus subscribers, who pay $20 per month. OpenAI was explicit that the 5-hour cap will not return to the $100 and $200 per month Pro plans. Enterprise and Education accounts use a separate credit-based system and are similarly unaffected.

That distinction matters. For individual professionals or small teams using Plus as a budget option for AI-assisted coding and autonomous workflow tasks, the restored cap is a real constraint to plan around. For organisations that have already moved to Enterprise or built internal tooling on top of the API, it changes nothing.

What This Means for Business

If you are on Plus: The free-flowing access you may have experienced over the past few weeks is gone. Heavy coding sessions or long-running agentic tasks will hit the 5-hour ceiling. You will need to spread intensive work across time, or upgrade your plan.

If you are evaluating ChatGPT for team use: This is a reminder that consumer-tier AI products are optimised for the average user, not for teams with consistent, high-volume workloads. Caps, quotas, and policy changes are part of the deal when you rely on shared infrastructure you do not control.

The broader pattern: This is the third time in 12 months that OpenAI has adjusted Plus usage limits in response to demand spikes. Each time, Enterprise and Pro customers were insulated. The gap between consumer AI tiers and enterprise-grade deployments is not just about raw capability, it is about reliability, predictability, and the absence of surprise policy changes that disrupt how your team works.

Businesses that depend on AI for core operations are learning this the hard way. Teams that built workflows around Plus-tier access now need to absorb a policy change that was announced with 24 hours’ notice.

That is not a knock on OpenAI specifically. It is the nature of shared cloud AI infrastructure. Organisations that need consistent, governed access to AI capabilities are better served by dedicated API access or purpose-built deployments where usage terms are fixed by contract, not subject to live product decisions.

Enterprise DNA works with businesses on exactly this kind of transition, helping them move from ad-hoc AI tool use to structured deployments where the rules of the game do not change without warning.

If your team is hitting limits on shared AI tools and needs a more reliable foundation for AI-assisted work, talk to us about what a custom deployment looks like.