ChatGPT 5-Hour Limit Is Back in 2026: What Changed for Plus, Codex & ChatGPT Work?
OpenAI has brought back its five-hour usage window for ChatGPT Plus users using Codex and ChatGPT Work after temporarily removing the short-term restriction. The weekly limit has not disappeared, which means many users now have two usage windows to think about again. Here is what actually changed, why the five-hour window returned, who it affects, and what we can expect next.
The five-hour limit does not mean ordinary ChatGPT stops working after five hours. The change mainly relates to agentic usage through products such as ChatGPT Work and Codex. Standard ChatGPT conversations and individual models can have separate usage limits.
ChatGPT's 5-Hour Usage Window Is Back
Over the past several weeks, regular users of OpenAI's more advanced agentic tools may have noticed something unusual inside their usage dashboard.
The familiar five-hour restriction disappeared.
For a temporary period, OpenAI allowed eligible users to work largely against their weekly allowance without having the same shorter five-hour usage window getting in the way.
For developers and other heavy users, this was a significant change.
Someone working intensively with Codex could continue coding for a longer uninterrupted session instead of reaching a short-term allowance while still having plenty of weekly usage remaining.
That temporary arrangement has now changed.
In late August 2026, OpenAI restored the five-hour usage allowance for ChatGPT Plus users across ChatGPT Work and Codex.
The weekly allowance continues to exist.
In practical terms, Plus users now need to pay attention to both the shorter five-hour allowance and their broader weekly allocation.
ChatGPT 5-Hour Limit Timeline
Why Did OpenAI Bring Back the 5-Hour Limit?
Whenever a technology company introduces another usage restriction, the immediate assumption is usually that the company simply wants people to use the product less.
The reasoning behind this change is more complicated.
OpenAI has pointed toward two major concerns: managing expensive computing demand and preventing users from consuming a large percentage of their weekly allowance during one heavy session.
1. Agentic AI Uses Much More Computing Power
A normal ChatGPT question might involve asking for an explanation, rewriting a paragraph, summarising some text, or solving a straightforward problem.
Agentic work can be very different.
Imagine asking Codex to understand a large application.
It may need to inspect dozens of files, understand how different components communicate, identify an error, modify multiple files, execute commands, check the results and continue reasoning when something fails.
That is far more demanding than generating a short text response.
The same applies to increasingly capable ChatGPT Work tasks that can involve documents, tools, research, code and multiple stages of work.
When thousands or millions of users are doing this simultaneously, the computing requirement becomes substantial.
A five-hour window gives OpenAI another way to distribute this heavy usage instead of allowing every user to consume most of their weekly capacity during the same short period.
2. It Stops Users From Accidentally Burning Their Weekly Allowance
This is probably the most interesting reason for bringing the limit back.
Removing the five-hour restriction gave users more freedom, but freedom also created a new problem.
A particularly demanding Codex or Work session could consume a surprising amount of the weekly allowance.
A developer could start working on Monday morning, run multiple complicated tasks, and later discover that a large percentage of the allowance intended to last until the next weekly reset had already disappeared.
The shorter limit acts as a pacing mechanism.
Does the 5-Hour Reset Give You More Weekly Usage?
No.
This is one of the most important things to understand about the new ChatGPT usage system.
The five-hour allowance and the weekly allowance work alongside each other.
A fresh five-hour window does not automatically restore your weekly allowance.
For example, imagine that you still have plenty of your weekly allocation remaining but you have exhausted your current five-hour allowance.
You may have to wait for the five-hour window to reset even though your weekly usage dashboard still shows remaining capacity.
The opposite can happen as well.
Your new five-hour window may begin, but if your broader weekly allowance has already been exhausted, the shorter reset does not magically give you another full week's worth of included usage.
The easiest way to understand the system is to imagine two separate gates.
| Usage Limit | What It Controls | What Happens When You Reach It? |
|---|---|---|
| 5-Hour Window | Controls concentrated short-term agentic usage. | You generally need to wait for the applicable short-term reset or use another available paid usage option. |
| Weekly Limit | Controls the total amount of included usage available over the broader weekly period. | You need to wait for your weekly allowance to reset or use additional credits where supported. |
Which ChatGPT Features Can Use This Allowance?
ChatGPT's usage system is becoming more complicated because different agentic products can now share usage allowances.
According to OpenAI's current documentation, supported agentic products can draw from the same allowance depending on your subscription and the products available to your account.
- Codex — OpenAI's coding agent and software-development environment.
- ChatGPT Work — longer-running work and deliverable creation inside ChatGPT.
- ChatGPT for Excel — supported spreadsheet-related agentic workloads.
- Workspace Agents — agentic workflows available to supported accounts.
Not every user has access to every feature.
What matters is the shared usage concept.
If two products use the same agentic allowance, heavy usage in one product may reduce the capacity you later see when you open another.
Does Normal ChatGPT Usage Have the Same 5-Hour Limit?
No, and this is where headlines can easily become misleading.
Saying that “ChatGPT has a five-hour limit” makes it sound like someone can use ChatGPT for five hours and then the entire service becomes unavailable.
That is not what this change means.
Standard ChatGPT conversations can have separate limits depending on the subscription, model and feature being used.
The five-hour change being discussed in August 2026 relates specifically to heavier agentic usage, particularly around ChatGPT Work and Codex.
Someone using ChatGPT mainly for writing, brainstorming, summaries, studying or regular questions therefore should not assume that ChatGPT will simply stop functioning every five hours.
How Many Messages Can You Send in Five Hours?
There is no reliable universal number.
This is another important difference between modern agentic AI and older message-based usage limits.
Twenty simple requests may require fewer resources than one extremely complicated task.
Usage can depend on factors including:
- The model being used
- The complexity of the request
- The amount of previous context
- The number and size of files involved
- How much code needs to be inspected
- Whether tools or commands need to run
- How long an agent continues working
- The amount of reasoning required
Consider two developers.
Developer A asks Codex to rename a variable inside one small file.
Developer B asks Codex to investigate an authentication bug across a large Flutter application and Laravel backend, reproduce the issue, modify several files, execute tests and verify the solution.
Both developers technically sent “one prompt.”
The amount of actual AI work behind those requests could be completely different.
This is why simply counting messages is becoming a poor way to understand ChatGPT's advanced usage limits.
What About ChatGPT Pro Users?
This is where OpenAI appears to be creating a clearer distinction between standard paid subscriptions and plans intended for heavy professional usage.
When the five-hour restriction was restored for Plus users, OpenAI indicated that the specific five-hour limit would remain disabled for its higher Pro subscription tiers for the coming months.
That does not mean that every feature on ChatGPT Pro has absolutely unlimited usage.
Different models, tools and services can still have their own policies, allowances and fair-use protections.
However, removing the short-term agentic restriction from higher Pro plans makes those subscriptions considerably more attractive to developers and professionals who routinely need long uninterrupted sessions.
This could become one of the most meaningful differences between ChatGPT Plus and higher-priced plans.
What About ChatGPT Business?
ChatGPT Business usage can vary depending on the type of seat and the organization's setup.
OpenAI has been introducing different levels of included agentic capacity, meaning users should not assume every Business account behaves exactly the same way.
If you use ChatGPT through an organization, the safest approach is to check the usage information provided inside your own workspace.
Why the 5-Hour Limit Could Actually Help Some Users
Any restriction naturally sounds negative.
But for casual or moderate users, the five-hour system has some practical advantages.
Your Weekly Allowance Becomes Harder to Waste
Without a short-term limit, one unusually heavy work session could consume a large amount of your weekly capacity.
The shorter window creates a natural stopping point before that happens.
Users therefore have less risk of opening ChatGPT on Monday, experimenting with multiple heavy projects and discovering on Tuesday that most of the week's included usage is already gone.
OpenAI Can Spread Demand Across the Day
Advanced AI models require expensive infrastructure.
If users are prevented from concentrating an extreme amount of agentic work into one short period, server demand becomes easier to distribute.
In theory, better demand management may also help OpenAI provide larger total allowances than it could if every subscriber were allowed completely unrestricted burst usage.
Usage Can Become More Predictable
For someone who only uses Codex for a few tasks each day, the five-hour limit may provide additional protection without significantly affecting their normal workflow.
The Biggest Problem With the 5-Hour Limit
Not everybody works evenly throughout the week.
This is where the five-hour system becomes frustrating.
Imagine a developer who works on their personal project every Saturday.
They may have almost their entire weekly allowance remaining.
Saturday is the one day when they have six or seven uninterrupted hours available to work.
A short-term limit can interrupt that session even though the account still has considerable weekly capacity.
Freelancers may face the same situation before a deadline.
A founder preparing a product release may suddenly need an intensive day of AI-assisted coding.
A researcher may want to complete one large analysis session rather than spreading it over several days.
All of these users may understandably ask:
“If I still have weekly usage remaining, why can't I decide when I want to use it?”
That will probably remain the strongest criticism of the five-hour system.
What Happens When You Reach the ChatGPT Limit?
OpenAI increasingly provides more than one path when users exhaust included usage.
Depending on your subscription and account, you may be able to:
- Wait until your five-hour allowance resets
- Wait until the weekly allowance resets
- Use purchased ChatGPT credits where supported
- Move to a subscription with a larger included allowance
- Use a less resource-intensive workflow where appropriate
Credits are particularly interesting.
They allow OpenAI to offer a middle ground between completely fixed subscription limits and forcing every heavy user to purchase a dramatically more expensive plan.
A person who exceeds the normal allowance once every few months may prefer purchasing extra credits instead of upgrading their subscription permanently.
How ChatGPT Plus Users Should Manage Usage
Most ordinary users should not dramatically change how they use ChatGPT.
If you regularly use Codex or ChatGPT Work, however, a few habits can help.
- Check the usage dashboard before starting a very large project. Do not assume that you still have capacity simply because you have not sent many prompts.
- Watch both the five-hour and weekly allowances. One meter can become restrictive even while the other still has capacity.
- Use agentic workflows when they provide real value. A small question does not always require a long autonomous task.
- Break extremely large jobs into clear stages. This can also improve results because the agent receives more focused instructions.
- Choose models intentionally. The most powerful option is not automatically the best option for every task.
- Keep unnecessary context out of large coding tasks. Large amounts of irrelevant context can make AI work more expensive and sometimes less focused.
What Should We Expect From ChatGPT Limits Next?
Nobody outside OpenAI can know exactly how its pricing and usage systems will evolve.
However, the latest changes give us several strong clues about the direction ChatGPT is heading.
1. ChatGPT Plans Will Be Separated by Compute, Not Just Features
In the earlier days of ChatGPT, subscription differences were relatively straightforward.
Pay for a higher plan and you might receive access to a better model, faster responses or more messages.
Agentic AI changes that equation.
One user might ask ChatGPT to explain a paragraph.
Another might ask an AI agent to spend several minutes investigating a software repository and performing dozens of internal steps.
Those users are not consuming comparable resources.
Future subscriptions are therefore likely to distinguish users increasingly through the amount and speed of included compute.
2. Credits Will Probably Become More Important
Credits make sense for both OpenAI and users.
They allow occasional heavy users to pay for extra capacity without changing their entire subscription.
As ChatGPT becomes capable of increasingly expensive autonomous tasks, pay-as-you-go capacity is likely to become a larger part of the product.
3. Message Counts Will Matter Less
The future of ChatGPT limits probably will not revolve around simple statements such as “you can send 100 messages.”
Agentic systems may search the web, open files, execute code, inspect applications, perform analysis and continue working after the user sends a single instruction.
One prompt can therefore represent hundreds of internal operations.
Usage systems will increasingly need to account for the work being performed rather than only the number of messages.
4. Efficient Models Will Become More Valuable
Users often assume that the largest or most capable model should be selected for every problem.
That may become increasingly inefficient.
If a smaller model can complete a straightforward task successfully while consuming less of an allowance, choosing the right model becomes part of effective ChatGPT usage.
The best model will increasingly mean the model that provides enough capability for the task at an appropriate cost.
5. Better Usage Dashboards Will Be Essential
Shared allowances introduce another challenge: transparency.
If someone spends Monday using Codex and Tuesday using ChatGPT Work, they need an easy way to understand why their remaining capacity changed.
Users will increasingly expect:
- Clear reset dates and times
- Remaining usage percentages
- Separate five-hour and weekly indicators
- Better explanations of expensive tasks
- Credit balances
- Estimated usage before very expensive operations
As pricing becomes more compute-based, clear usage information becomes just as important as model performance.
Is the Return of the ChatGPT 5-Hour Limit Good or Bad?
The answer depends almost entirely on how you use ChatGPT.
For someone who occasionally asks Codex to investigate code or uses ChatGPT Work for a few projects each week, the five-hour restriction may barely be noticeable.
In fact, the additional limit could help that person avoid accidentally consuming an entire week's allowance.
For a professional developer who intentionally wants to spend an entire day working with an AI coding agent, the experience is very different.
Having plenty of weekly usage remaining is not especially useful when a shorter limit interrupts the work session.
Neither perspective is unreasonable.
The important development is what the change tells us about ChatGPT itself.
ChatGPT is no longer simply a website where people type questions and receive text responses.
It is increasingly becoming a platform where AI can perform substantial amounts of real work.
And once AI starts doing work rather than simply generating messages, measuring and pricing that work becomes much more complicated.
Frequently Asked Questions About the ChatGPT 5-Hour Limit
Is the ChatGPT 5-hour limit really back?
Yes. OpenAI restored the shorter five-hour agentic usage window for ChatGPT Plus users across products including ChatGPT Work and Codex after temporarily removing it.
When did OpenAI bring the five-hour limit back?
The restoration was announced in late August 2026, with the change appearing for Plus users around August 25, 2026.
Does ChatGPT stop working after five hours?
No. The five-hour window discussed here applies to supported agentic usage rather than meaning the entire ChatGPT service stops working after five hours.
Is the five-hour limit separate from the weekly limit?
Yes. The shorter five-hour allowance and broader weekly allowance can both apply to eligible agentic usage.
Does a new five-hour window restore my weekly ChatGPT limit?
No. Resetting the short-term allowance does not automatically restore weekly usage that has already been consumed.
Why did OpenAI restore the five-hour limit?
One reason is to better distribute computing demand. The shorter window can also prevent users from accidentally consuming a large percentage of their weekly included usage during one intensive session.
Does normal ChatGPT chatting use this five-hour limit?
Not in the simple way the phrase suggests. Regular ChatGPT models and features can have their own limits. The August 2026 five-hour change primarily concerns heavier agentic usage such as Codex and ChatGPT Work.
How many Codex requests can I make every five hours?
There is no universal request count because usage varies according to model choice, task complexity, context size, tools and how much work the agent performs.
Can I continue using Codex after reaching the limit?
Depending on your plan and account, you may need to wait for the relevant reset or use additional paid credits where OpenAI makes them available.
Where can I check my ChatGPT usage limit?
Open ChatGPT's usage information or the relevant Codex usage dashboard. The available usage indicators can differ depending on your subscription and features.
Final Thoughts
The return of ChatGPT's five-hour agentic usage window is not quite as dramatic as saying “ChatGPT has a five-hour limit,” but it is still an important change.
During the temporary period without the short-term restriction, users had much greater freedom over when they consumed their weekly allowance.
That was particularly useful for people who prefer long, concentrated coding and productivity sessions.
The downside was that an intensive session could consume a surprisingly large part of a weekly allowance before the user realised what had happened.
OpenAI has now chosen a compromise.
Plus users retain their broader weekly allowance, but the five-hour window limits how rapidly some of that included agentic capacity can be consumed.
Casual ChatGPT users may barely notice.
Developers, researchers, founders, freelancers and other people who rely heavily on Codex or ChatGPT Work are much more likely to feel the difference.
But the most interesting part is not the five-hour number itself.
The bigger story is how ChatGPT usage is evolving.
As AI moves from answering questions to reading projects, writing software, researching information and performing multi-step work, simple message limits no longer describe what users are actually consuming.
That means the future of ChatGPT subscriptions will probably be shaped increasingly by compute, agentic usage, model efficiency and credits.
The return of the five-hour window is one of the clearest signs of that transition so far.
This article was prepared using currently available OpenAI usage-limit documentation and public information available in August 2026. ChatGPT plans, model limits, Codex limits, credits and agentic usage policies can change. Always check the Usage section of your own ChatGPT account for the limits currently applied to your subscription.
