Codex’s Five-Hour Limit Returns on Plus, While Pro Stays Exempt for Months: Is That Reason Enough to Upgrade?
On August 25, 2026, Thibault Sottiaux, who works on Codex and ChatGPT at OpenAI, announced that the five-hour limit would return for Codex and ChatGPT Work on the Plus plan. At the same time, the limit will remain disabled for the $100-per-month Pro 5x plan and the $200-per-month Pro 20x plan for the next few months.
The important difference is not simply how much usage each plan includes, but whether a long session can continue without a five-hour interruption.
For me, the absence of a five-hour limit is a compelling reason to choose Pro. Avoiding an interruption in the middle of focused work may matter more than the larger allowance itself.
The five-hour limit returns on Plus, while Pro stays exempt for now
For the individual plans named in Sottiaux’s announcement, the change can be summarized as follows.
| Plan | Monthly price | Five-hour limit |
|---|---|---|
| Plus | $20 | Returns on August 25, 2026 |
| Pro 5x | $100 | Remains disabled for the next few months |
| Pro 20x | $200 | Remains disabled for the next few months |
OpenAI’s pricing documentation describes the standard five-hour window, while Sottiaux announced this temporary Pro treatment on X. Check what currently applies to your account in the Codex usage dashboard or with /status in the CLI.
The five-hour and weekly limits serve different purposes
The five-hour limit controls short-term usage pace; the weekly limit measures usage across the whole week. They serve different purposes.
You may still have room in your weekly allowance yet hit the five-hour limit during a day of research, implementation, and testing. You then need to switch tasks, use credits, or wait for the window to reset.
What remains disabled on Pro is the five-hour limit. The practical difference is how far you can take one piece of work without being forced to stop.
That may sound minor when you submit only short requests. With longer Codex jobs, the freedom to continue directly affects the workflow.
Why OpenAI is bringing the five-hour limit back to Plus
Sottiaux gave two reasons for restoring the limit on Plus.
- The five-hour window helps smooth demand on OpenAI’s compute capacity, making it easier to keep the weekly allowance generous.
- Plus users tend to include more casual and newer users, who may accidentally consume an entire week’s allowance and then be confused by the result.
This is reasonable for someone who uses Codex for a few focused sessions each week. For someone who works for several hours until a natural stopping point, the same limit can disrupt the workflow. That is the practical distinction OpenAI is temporarily drawing between Plus and Pro.
Without the five-hour limit, I want to use Codex more
I described the limit-free period in a Japanese-language post on our sister site, Delaymania. Once I stopped watching the five-hour window, I began spending more time in Codex than in Claude Code.
I previously relied more on Claude for planning and zero-to-one work. As I learned GPT-5.6 Sol, Codex also began producing ideas and implementations that worked well for me. I prefer Codex’s output for both code and prose, so fewer interruptions naturally lead me to open it more often.
This is a personal judgment, not an objective benchmark. Claude still makes a useful second reviewer for Codex output, and I do not want to abandon the division of labor between Claude and Codex.
When both tools are capable, I start with the one less likely to stop mid-job. Model intelligence dominates AI comparisons, but the ability to finish today’s work matters just as much.
Pro should sell continuity, not only more usage
The best outcome would be for OpenAI to keep this as a clear, lasting reason to choose Pro rather than a temporary promotion.
Pro 5x and Pro 20x offer five and twenty times the Plus usage. Those numbers describe capacity, but not which plan a person actually needs.
“Long sessions are less likely to stop every five hours” is easier to understand. It matters to people building websites or apps, writing articles, and organizing many documents. At professional-plan prices, users are paying for both capacity and a more predictable working day.
OpenAI may not be able to promise this permanently if it affects compute demand or stability. Even so, it could make the experience more predictable.
- Keep Pro free of the five-hour limit by default, with temporary adjustments only during exceptional demand
- Give users advance notice and a clear end date when the limit must return
- Show whether the five-hour limit applies and when it will reset in a clear, consistent place
- Provide practical guidance for deciding whether Pro 5x or Pro 20x fits a user’s actual pattern
Once you work without this interruption, its return feels much more disruptive. OpenAI could turn a temporary incentive into a durable reason to stay subscribed.
Who should consider upgrading to Pro now?
Pro is not right for everyone. At $100 per month or more, the first question is where the five-hour limit actually interrupts your work.
| How you use Codex | How to think about the plan |
|---|---|
| A few short tasks each week | Plus is probably enough |
| Light use most of the time, with occasional deadline spikes | Compare the cost of Plus with additional credits against a Pro subscription |
| Long sessions with repeated back-and-forth on one job | Consider Pro 5x, including the cost of time lost to interruptions |
| Pro 5x still does not provide enough total usage | Consider Pro 20x after reviewing your actual usage dashboard |
Track two things for one week.
- How many times the five-hour limit stops an active task
- How much time is lost waiting or moving the work to another service
If the limit repeatedly interrupts paid work, Pro’s continuity may be valuable. If it happens occasionally, compare Plus credits with Pro’s much higher monthly price.
Our article on who actually needs a premium AI plan also evaluates expensive subscriptions by working time rather than features alone.
“The next few months” is not a permanent promise
OpenAI did not provide an end date. Sottiaux said the limit would remain disabled on Pro for “the upcoming months,” not permanently.
- Pro usage itself is not unlimited
- No end date has been announced for this exemption
- Codex and ChatGPT Work share the same usage allowance
- The standard pricing documentation and temporary operating policy may not always show the same thing at the same time
It is too early to plan a full year assuming Pro will never have a five-hour limit. Before upgrading, ask whether several uninterrupted months justify the cost now, and keep checking the usage dashboard.
Make the absence of a five-hour limit a standard Pro benefit
The five-hour limit is returning on Plus, while Pro 5x and Pro 20x keep it disabled for the next few months. The plans now differ in both capacity and continuity.
Plus suits people comfortable with shorter windows. Pro suits those who want to continue until the job reaches a natural stopping point. Choosing between those working styles has real value.
Without the five-hour limit, I want to use Codex more. I hope OpenAI makes that a lasting reason to choose Codex for professional work.
Sources
- Tibo (Thibault Sottiaux) on X: the five-hour limit on Plus and the temporary Pro exemption
- OpenAI: ChatGPT Work and Codex pricing and usage limits
- Delaymania: Codex is much easier to use without the five-hour limit (Japanese)
Plans, prices, and usage limits were checked on August 25, 2026. Account-specific limits and temporary policies may change, so check the Codex usage dashboard and OpenAI’s pricing documentation before changing plans.
