As checked on October 18, 2026, OpenAI lists five-hour usage estimates for Plus and Standard Business, while Pro currently has no five-hour limit. Weekly limits may still apply. Use your Usage Dashboard for the exact limits and reset times attached to your account.
What the 5-hour window actually is§
As checked on October 18, 2026, OpenAI lists five-hour usage estimates for Plus and Standard Business, while Pro currently has no five-hour limit. Weekly limits may still apply. Use your Usage Dashboard for the exact limits and reset times attached to your account.
The public documentation does not describe a universal per-message replenishment formula. Avoid treating phrases such as "usage ages out continuously" as established fact unless the product itself shows that behavior for your account.
The July 2026 removal — and what came after§
On July 13, 2026, OpenAI temporarily removed the 5-hour limit for Codex, a move widely read as a competitive response in the coding-agent race. It landed in the middle of the densest reset spree on record — 7M-to-10M-user milestones announced with back-to-back global refills — and the team confirmed days later that the removal would stay for the time being.
Its status has changed before and can change again. The way to know what's true today is the official signals: the team's announcements (tracked in the reset history) and your own /status readout, which reflects whatever policy is currently in force.
5-hour vs weekly: which one is stopping you§
Quick triage when you're throttled: open Usage and read both the five-hour and weekly indicators plus their reset times. Do not diagnose the limiting window from elapsed time alone.
Model choice changes the estimate: OpenAI says simpler tasks may consume only a fraction of the allowance while complex codebases, long-running tasks and extended sessions consume more. Use the model and reasoning level that fit the task rather than assuming a fixed message cost.
How to check its current status§
Open the Usage Dashboard for exact reset times, or run /status in the Codex CLI for remaining limits. For the separate public-event layer, the live radar tracks official announcements and compares the current gap with history.
One thing worth remembering from the removal: these policies are tools the team adjusts in response to load and competition. The stable strategy is the one this site preaches — understand the current rule, plan heavy work around refills, and never burn account health on bypass tricks.
Reading your window state like a mechanic§
Two indicators decide your next move: remaining five-hour allowance and remaining weekly allowance. If only the short allowance is low, schedule work after the reset time shown in Usage or switch to a lighter task. If both are low, the weekly timestamp and any official reset option shown for your account matter most.
Do not infer a hidden algorithm from one recovery. Record the displayed reset time, what model you used and the task type. A few first-party observations are useful for personal planning; they still do not establish a universal rule.
Making one window last the whole day§
Three pacing patterns are safe without guessing the reset algorithm. Time-box: place expensive work after the reset time displayed by your account. Model-fit: reserve heavyweight models and high reasoning for tasks that need them. Scope control: reduce needless context and repeated retries by giving a clear goal and a focused file set.
None of this raises your ceiling; it removes self-inflicted walls. The consistent report from community threads is that pacing alone recovers a meaningful share of "lost" quota on the same tier — before any upgrade math enters the picture.
A historical change is not a current entitlement§
The July removal belongs in the dated event archive. It does not establish the limit of every plan today. The current pricing page explicitly distinguishes Pro from Plus and Standard Business.
When reading an old screenshot, note the date and plan. When planning current work, read your current account. Keep the two pieces of evidence separate.
A throttle-moment decision tree§
The moment you're throttled, ask three questions. One: which allowance is low? Read Usage or /status. Two: when does the account say it resets? Use that exact timestamp in the timer, not a guessed weekday. Three: can the work change? Defer the heavy task, narrow its scope or use a lighter model if the task allows.
Only after those checks should you consider an eligible reset option shown in your account. A public reset announcement may add context, but an archive-derived pressure gauge alone is not a reason to spend or burn quota.
FAQFAQ
Is the Codex 5-hour limit gone for good?
As checked on October 18, 2026, OpenAI lists five-hour usage estimates for Plus and Standard Business, while Pro currently has no five-hour limit. Weekly limits may still apply. Use your Usage Dashboard for the exact limits and reset times attached to your account.
How is the 5-hour limit different from the weekly limit?
OpenAI estimates included usage per five-hour period and notes that weekly limits may also apply. Your Usage Dashboard shows the exact account-specific limits and reset times.
What time does the 5-hour allowance reset?
Use the exact time shown in your Usage Dashboard. OpenAI does not publish one universal reset clock that this site can apply to every account.
Why do I get throttled even with weekly usage left?
If your account displays both limits, one can be exhausted while the other has remaining capacity. Check the displayed label and reset time. Pro currently has no five-hour cap.
Does the 5-hour limit apply to cloud tasks too?
Local messages and cloud tasks share plan usage. The five-hour period applies only where your plan has that limit; Pro currently has no five-hour cap.
Does logging out or switching devices reset the window?
No. The window follows your account, not your session or device. Re-logging changes nothing about usage that's still inside the window.
Is there any official way to raise the 5-hour ceiling?
Only the legitimate levers: a higher subscription tier, lighter models for routine work, and pacing. Everything marketed as a "limit bypass" violates the terms and risks the account.
Does stopping use make the allowance refill faster?
OpenAI does not document a bonus refill rate for inactivity. Stop new usage if you want to conserve remaining capacity and follow the reset time shown in your account.
What does 'local and cloud share the window' mean in practice?
One account-level pool: a cloud batch you fired off still counts while you keep prompting locally, so a long-running cloud task plus heavy local work is the fastest way to empty the 5-hour window. Budget them together, not as separate budgets.