OpenAI just put the five-hour gate back on ChatGPT Plus. After weeks of looser Codex and ChatGPT Work access, the cheaper tier is back to rationing the one thing developers wanted most: long, uninterrupted runs.
If you're a ChatGPT Plus subscriber, the free ride is over. For several weeks you could run Codex and ChatGPT Work without watching the short-window meter so closely. Not any more. As 9to5Mac reported, Thibault Sottiaux, OpenAI's engineering lead for Codex and ChatGPT, said on X that OpenAI would bring back the five-hour limit for Plus accounts across Codex and ChatGPT Work on August 25. The timing matters. OpenAI temporarily lifted the five-hour Codex limit in July for Plus, Pro and Business users, according to an OpenAI Developer Community post from support staff dated July 28. That left the weekly limit in place, but removed the short-window cap that could stop a coding session while you still had weekly usage left. For anyone using Codex as a working tool rather than a demo, that was the difference between staying in flow and waiting for a reset.
Six weeks is not long. It was long enough for developers to build habits around it.
The Limit Came Back For A Reason #
Sottiaux's stated reason was compute, not discipline. In the August 24 post quoted by 9to5Mac, he said the five-hour limit helps OpenAI smooth demand on its compute and keep the Plus plan generous on weekly usage. He also said some Plus users were burning through a whole week's allowance in one sitting and then getting confused by the experience. That's a plain product problem. It is also a capacity problem.
OpenAI Now Lets You Link Multiple Gmail Accounts to ChatGPT OpenAI confirmed on August 28, 2026 that ChatGPT and Codex now support linking multiple Gmail, Google Calendar, and Google Contacts accounts in one conversation, ending a year-old limitation that forced users to pick a single Google account. The update is live on Plus, Pro, Business, and Enterprise plans, though workspace admins can still restrict... - how to link multiple Gmail accounts to ChatGPT - ChatGPT multiple Google account support now available
OpenAI's own help center now says usage from Codex, ChatGPT Work, ChatGPT for Excel and Workspace Agents draws from the same agentic usage and credit pool when those features are available on your plan. A Codex task can use very different amounts of allowance depending on the model, the size of the codebase and where the task runs. Small scripts barely touch the pool. Long-running refactors can chew through it fast.
That is the hard part for Plus users. You don't experience Codex usage as tokens, credits or a table on a pricing page. You experience it as a tool stopping in the middle of real work. That's an easy line in a meeting. It feels very different when you're the one trying to wrestle a messy pull request into shape before the day ends, deadline breathing down your neck.
OpenAI had already spent August cleaning up the usage-drain mess. Sottiaux posted on August 8: he'd reset usage limits for paid ChatGPT Work and Codex users. Later in the month, he pointed to specific inefficiencies: images in long sessions with multiple compactions, high p95-plus usage tied to Computer History, and a conversation-title feature that was draining more usage than intended. That is not abstract capacity talk. Those are background product costs landing inside a user's quota.
Waste, basically.
On July 29, Sottiaux also said OpenAI had made improvements after users complained that GPT-5.6 Sol was using Codex limits faster than expected. He wrote that OpenAI expected typical Sol usage to last around 18% longer after those changes, while also saying the five-hour limit would be restored after the temporary . The OpenAI rate card lists GPT-5.6 Sol at $4 per million input tokens and $20 per million output tokens for token-based ChatGPT Work and Codex usage. Every extra tool call has a cost somewhere.
Pro Users Get A Different Deal #
The sharper part of the announcement is what did not happen to Pro. Sottiaux said OpenAI would keep the five-hour limit disabled for the $100 and $200 Pro subscriptions for the coming months. Plus users get the gate back. Pro users keep the longer runway, at least for now.
Frankly, that is the real pricing message. OpenAI is not saying Codex cannot support long sessions. It is saying long sessions now belong more clearly to the higher-paying plans. OpenAI's Codex pricing page describes Pro as built for longer, higher-intensity sessions and says Pro users get five times or 20 times more usage than Plus, depending on the tier. That is not subtle.
OpenAI resets Sol usage limits and fixes the efficiency gap that caught power users off guard
OpenAI reset usage limits for all ChatGPT Work and Codex subscribers and rolled out efficiency fixes for GPT-5.6 Sol after the model's agentic behavior drained quotas faster than expected. Product lead Tibo Sottiaux said typical usage should now last around 18% longer, and acknowledged the company should have caught the problem sooner. - openai sol usage limits reset - gpt 5.6 sol draining allowances
If you're paying $20 a month and using Codex for occasional help, the change may barely touch you. If you're using it for actual development work, you need to treat the five-hour window as part of the workflow again. Start the heavy jobs when you can stay with them. Save the lighter fixes for later. Don't pretend the weekly allowance is the only limit that matters. OpenAI remains one of the best-funded AI companies in the world, and even it is making customers feel the shape of the compute bill. GPU capacity is not a footnote here. It is the product.
For developers who rearranged their days around looser Codex access, the practical point is simple: budget your five-hour windows again. OpenAI has decided they are finite. Also read: G7 Bond Yields Surge to Highest Level Since 2004 as Debt Bills Balloon • SpaceX Is Building a Turbine Blade Factory to Escape the AI Power Crunch • Copper Just Hit a Record Above $14,800 a Tonne and AI Is Partly to Blame