ChatGPT usage limits are back on the menu. Late on Monday 24 August 2026 (01:16 UTC on 25 August), OpenAI’s engineering lead for Codex and ChatGPT, Thibault “Tibo” Sottiaux, announced on X that the rolling 5-hour usage window would return the next day for ChatGPT Plus accounts across ChatGPT Work and Codex. “Tomorrow we will bring back the 5h limit for Plus accounts across ChatGPT Work and Codex,” he wrote. “I had mentioned this a while ago, but then postponed it.”
The detail most headlines garbled: the Pro tiers escape. Sottiaux was explicit that the 5-hour ChatGPT usage limit stays switched off for both Pro subscriptions “for the upcoming months”. And this is not a cap on ordinary chat messages — it governs the agentic products, ChatGPT Work and Codex, where a single session can burn through an enormous token budget.
This article unpacks exactly what the reinstated ChatGPT usage limit covers, the six-week experiment that preceded it, who is exempt and for how long, how the numbers compare with Claude and Gemini, and what Plus subscribers can actually do when they hit the wall.
Table of contents
- What the New ChatGPT Usage Limit Actually Covers
- Why OpenAI Reintroduced the 5-Hour ChatGPT Usage Limit
- The Six-Week Timeline Behind the ChatGPT Usage Limit U-Turn
- Plus vs Pro: Who Escapes the ChatGPT Usage Limit
- How the ChatGPT Usage Limit Compares With Claude and Gemini
- What Plus Subscribers Can Do When They Hit the Cap
- What the ChatGPT Usage Limit Means for Your Business
- ChatGPT Usage Limit FAQ
- References
What the New ChatGPT Usage Limit Actually Covers
The reinstated ChatGPT usage limit is a rolling 5-hour window that sits underneath the weekly allowance. The weekly limit never went away; what returns on Tuesday 25 August (US time) is the second, shorter meter that stops one marathon session from draining a whole week of usage.
The scope: ChatGPT Work and Codex, not chat
The ChatGPT usage limit applies to ChatGPT Work — the workspace-agents mode OpenAI launched in July — and to Codex, the coding agent. Regular ChatGPT conversations run on a separate metering system and are untouched by this change. Anyone quoting generic chat-message figures at this news is mixing up two different systems.
Who is capped and who is not
Only ChatGPT Plus, the $20-per-month tier, gets the 5-hour ChatGPT usage limit back. Sottiaux’s post carved out the premium tiers in plain words: “We are for the upcoming months keeping the 5h limit not enabled for Pro $100 and Pro $200 subscriptions.” Enterprise and Edu customers run on a separate credit-based system and are unaffected. The Business plan — which was included when the limit was removed in July — went unmentioned in the reinstatement post.
| Plan | Price | 5-hour limit from 25 Aug | Weekly limit |
|---|---|---|---|
| ChatGPT Plus | $20/month | Reinstated | Yes (never removed) |
| Pro 5x | $100/month | Off “for the upcoming months” | Yes |
| Pro 20x | $200/month | Off “for the upcoming months” | Yes |
| Business | $20/user/month (annual) | Not stated in the announcement | Yes |
| Enterprise / Edu | Custom | Separate credit-based system | Credit-based |
How the ChatGPT usage limit meters stack
With the change live, a Plus account is measured twice at once. The 5-hour window refills on a rolling basis; the weekly cap sits above it. Hitting either one pauses agentic work until a reset arrives, a banked reset is spent, credits are purchased, or the account is upgraded. That is precisely the structure Codex ran from 2025 until mid-July — which is why “reintroduce” is the accurate verb.
Why OpenAI Reintroduced the 5-Hour ChatGPT Usage Limit
Sottiaux gave two reasons, both in the announcement itself — and neither is about punishing heavy users.
Reason one: smoothing the compute curve
“This is necessary as (a) the 5h limit allows us to smoothen the load on our compute, allowing to keep the plan generous in terms of weekly usage,” he wrote. Spiky demand is expensive: a fleet sized for everyone’s simultaneous marathon sessions sits idle the rest of the time. A rolling ChatGPT usage limit flattens the peaks, and OpenAI’s argument is that the flattening is what pays for a generous weekly allowance in the first place.
Reason two: protecting casual subscribers from themselves
The second reason is stranger and more human: “(b) users on the Plus plan are relatively casual and new users, but then also just accidentally eat through their whole weeks usage and then are confused, making it not a great experience.” In other words, without the short meter, a newcomer could point an agent at a task on Monday morning and discover by lunchtime that their entire week was spent. The 5-hour ChatGPT usage limit acts as a circuit breaker before the weekly one becomes a brick wall.
The backdrop: agents are hungry
Both reasons trace back to the same shift. A modern agentic session is not a chat with a large language model — it is hours of autonomous tool calls, code execution and long-context reasoning, and token consumption scales accordingly. OpenAI launched ChatGPT Work on 9 July alongside the GPT-5.6 family, and demand roughly doubled within 48 hours. We covered the model family behind that surge in our complete guide to GPT-5.6 Sol.
The Six-Week Timeline Behind the ChatGPT Usage Limit U-Turn
The reinstatement only makes sense against the unusual six weeks that preceded it — a period in which OpenAI removed the limit, reset everyone’s meters twice, promised a return date, and then blinked.
| Date | Event |
|---|---|
| 3 June 2025 | Codex access opens to ChatGPT Plus subscribers |
| 30 October 2025 | Codex credits launch: 1,000 credits for $40, plan usage consumed first |
| April 2026 | Codex metering switches from messages to tokens |
| 9 July 2026 | ChatGPT Work and the GPT-5.6 family (Sol, Terra, Luna) launch; demand roughly doubles in 48 hours |
| 12 July 2026 | 5-hour limit temporarily removed for Plus, Pro and Business; all usage reset; Sol optimisations add ~10% more usage |
| 29 July 2026 | Second full reset; Sol usage stretched ~18% further; limit’s return announced for 30 July — then postponed |
| 25 August 2026 | The 5-hour limit returns for Plus accounts only; Pro tiers stay exempt |
| 31 August 2026 | GPT-5.4 and GPT-5.4 mini retire from Codex |
The July holiday from the meter
When the limit came off on 12 July, Sottiaux framed it as pure upside: “No nerfing, only good stuff!” Plus, Pro and Business subscribers spent roughly six weeks governed only by the weekly cap, while OpenAI shipped inference optimisations that stretched Sol usage by around 10% and then a further ~18%. On 29 July he insisted “we have not reduced usage on any subscription plans” and pencilled in the limit’s return for the next day — a date that quietly slipped.
Why the reversal reads as a stress test
Read charitably, the whole window looks like a live capacity experiment: remove the short meter, watch what unconstrained agentic demand does to the fleet, then decide where the throttle must sit. The answer, six weeks later, is that the throttle sits on Plus. It is a familiar rhythm for OpenAI-watchers — the company also spent August recalibrating its safety governance, as we examined when the OpenAI preparedness team was disbanded.
Plus vs Pro: Who Escapes the ChatGPT Usage Limit
The exemption is the sharpest edge of this story. A $20 subscriber and a $200 subscriber now live under structurally different rules, not just different quantities.
What each tier actually gets per window
OpenAI’s official pricing documentation expresses Codex and ChatGPT Work allowances as local messages per 5-hour window, with wide model-dependent ranges. Plus spans 10 to 2,000 local messages per window depending on the model; Pro 5x runs five times that at 50 to 10,000; Pro 20x runs twenty times Plus at 200 to 40,000. Business matches Plus’s range on a per-user basis.
| Plan | Price | Local messages per 5-hour window | Multiplier vs Plus |
|---|---|---|---|
| ChatGPT Plus | $20/month | 10–2,000 (model-dependent) | 1× |
| Pro 5x | $100/month | 50–10,000 | 5× |
| Pro 20x | $200/month | 200–40,000 | 20× |
| Business | $20/user/month (annual) | 10–2,000 per user | 1× |
The ChatGPT usage limit multipliers make the gap visual. Each bar below encodes the tier’s allowance multiplier relative to Plus, as stated in OpenAI’s own plan naming — Pro 5x is five times Plus, Pro 20x is twenty times.
A ChatGPT usage limit exemption with no end date
“For the upcoming months” is the only duration OpenAI has put on the Pro carve-out. There is no published end date, and no commitment that the exemption survives the next demand spike. Anyone budgeting a workflow around uncapped Pro sessions is relying on a courtesy, not a contract.
The fairness argument brewing underneath
The two-rule structure invites an obvious complaint: the subscribers most likely to be throttled mid-task are the ones paying least, while the heaviest users the compute argument would target are exempt. OpenAI’s counter is embedded in reason (b) — Plus users are the ones who accidentally strand themselves — but the optics of capping the $20 tier while sparing the $200 tier will do their own work.
How the ChatGPT Usage Limit Compares With Claude and Gemini
The strangest thing about the 5-hour ChatGPT usage limit is how unremarkable it has become. All three frontier labs now ration their premium AI tools with some version of the same clock.
Anthropic got there first
Anthropic announced weekly caps for Claude Pro and Max in July 2025, layered on top of the 5-hour rolling resets it already ran, with enforcement from 28 August 2025. Claude Pro at $20 targets roughly 40 to 80 hours of Sonnet 4 per week; Max at $100 spans 140 to 280 hours of Sonnet plus 15 to 35 hours of Opus; Max at $200 reaches 240 to 480 hours of Sonnet plus 24 to 40 hours of Opus. Anthropic said fewer than 5% of subscribers would ever notice.
Google runs the same clock
Google’s own help documentation for Gemini states the pattern plainly: “Your limit refreshes every 5 hours until you reach your weekly limit.” Google AI Plus doubles the standard allowance, AI Pro quadruples it, and AI Ultra runs 5 to 20 times Pro depending on the feature.
| Provider | Limit structure | Since |
|---|---|---|
| OpenAI (ChatGPT Work / Codex) | 5-hour rolling window + weekly cap; Plus capped, Pro tiers exempt for now | Reinstated 25 August 2026 |
| Anthropic (Claude) | 5-hour rolling resets + weekly caps on Pro and Max | Enforced 28 August 2025 |
| Google (Gemini) | Compute-based limit refreshing every 5 hours, under a weekly ceiling | Current policy |
Convergence tells you where the costs are
Three rivals independently arriving at a 5-hour meter is not a coincidence — it is the shape of the underlying economics showing through every pricing page. Flat-rate subscriptions and unbounded agentic compute cannot coexist; the window is the compromise every lab has landed on. The same economics drive per-token API pricing, which you can explore with our AI token cost calculator.
What Plus Subscribers Can Do When They Hit the Cap
Hitting the ChatGPT usage limit mid-task is annoying but no longer a dead end. OpenAI has built four escape routes, and choosing the right one is mostly arithmetic.
Wait, bank, buy or upgrade
The free option is patience: the 5-hour window refills on its own. Second, banked resets — OpenAI’s mechanism for crediting back interrupted windows — can be spent to restart a session. Third, credits: 1,000 credits cost $40, remain valid for twelve months, and are only drawn down after the plan’s included usage is exhausted. Fourth, an upgrade to Pro 5x or Pro 20x currently buys out of the 5-hour meter entirely.
What credits actually buy
The official rate card prices GPT-5.6 Sol at 100 credits per million input tokens, 10 per million cached tokens and 500 per million output tokens. Terra runs cheaper at 50/5/300, and the lightweight Luna at 5/0.5/30. GPT-5.4, at 62.5/6.25/375, retires from Codex on 31 August — so credit budgets built around it need rebasing this week.
Early reaction to the ChatGPT usage limit
The early reaction suggests the cap will bite some workflows hard. Within hours of the announcement, a thread titled “Bring back the 5h limit for Plus accounts across ChatGPT Work and Codex” appeared on OpenAI’s developer forum, with one user writing: “With the 5h limits gone I have been able to get so much work done” and asking OpenAI to “reconsider the decision or at the very least consider daily limits.”
Others who subscribed during the uncapped window are calling the reversal a bait-and-switch — an echo of the wider trust debates around OpenAI’s product changes, from ChatGPT ad personalization to feature rollouts like the August ChatGPT quiz and reservations update.
What the ChatGPT Usage Limit Means for Your Business
For a business running real work through agentic AI, the reinstated meter is a planning input, not a catastrophe. The right response depends on how bursty your usage is.
Audit your burst profile first
A team that runs steady, modest agent tasks all week may never touch the 5-hour ceiling — the weekly cap was always the binding constraint. The teams at risk are the ones with deadline-shaped demand: a monthly reporting crunch, a release-day code sprint, a migration weekend. Map when your heaviest sessions happen before deciding whether the ChatGPT usage limit changes anything for you.
Price the four options against downtime
The arithmetic is straightforward. If a stalled agent costs an hour of a developer’s time, a $40 credit pack pays for itself the first time it is used. If the squeeze is weekly rather than occasional, the $100 Pro 5x tier — currently exempt from the ChatGPT usage limit as well as five times larger — is the cleaner fix. And if agents are load-bearing for revenue, price in a second provider: a quota you do not control is a single point of failure.
The scale behind these decisions keeps growing: OpenAI reported 900 million weekly active users and 50 million paying subscribers in February 2026, up from 800 million weekly users the previous October and 400 million in February 2025.
Treat exemptions and quotas as weather
The deeper lesson of the six-week saga is that subscription AI allowances are now dynamic. OpenAI changed the rules three times in six weeks, each time via a social media post from one engineering lead. Any business process that breaks when a quota moves needs either an upgrade path, a credit buffer, or a fallback provider — because the ChatGPT usage limit you plan around today is explicitly not guaranteed tomorrow.
ChatGPT Usage Limit FAQ
When does the ChatGPT usage limit come back?
From Tuesday 25 August 2026, US time. The announcement was posted at 01:16 UTC on 25 August — Monday evening in the United States — and said the change lands “tomorrow”.
Does the new ChatGPT usage limit affect Pro subscribers?
No. Both Pro tiers — Pro 5x at $100 and Pro 20x at $200 — keep the 5-hour limit switched off “for the upcoming months”, in OpenAI’s words. Only Plus accounts are capped, and Enterprise and Edu run on a separate credit-based system.
Does it apply to normal ChatGPT conversations?
No. The reinstated window governs ChatGPT Work and Codex — the agentic products. Ordinary chat messages are metered separately and are not part of this change.
What happens when I hit the cap?
Agentic tasks pause until the rolling window refills. You can also spend a banked reset, buy credits (1,000 for $40, valid twelve months, used only after plan allowance), or upgrade to a Pro tier.
Why did OpenAI bring the limit back?
Two stated reasons: smoothing load on its compute fleet so the weekly allowance can stay generous, and stopping casual Plus users from accidentally consuming an entire week’s usage in one session.
Do other AI providers impose the same limits?
Yes — the 5-hour window is now the industry norm. Anthropic runs 5-hour rolling resets plus weekly caps on Claude Pro and Max, and Google’s Gemini limits refresh every 5 hours beneath a weekly ceiling.
References
Thibault Sottiaux Announcement of the Returning 5-Hour Limit (X, via Xcancel)
OpenAI Restores 5-Hour Codex and Work Limits for ChatGPT Plus Users (9to5Mac)
ChatGPT Developer Pricing and Usage Documentation (OpenAI)
Using Codex With Your ChatGPT Plan (OpenAI Help Center)
Using Credits for Flexible Usage in ChatGPT (OpenAI Help Center)
Bring Back the 5h Limit for Plus Accounts (OpenAI Developer Community)
Codex Plus 5-Hour Limit Returns for ChatGPT Work (explainx)
OpenAI Restores 5-Hour Usage Limit for ChatGPT Plus Users (TechNews)
Anthropic Unveils New Rate Limits to Curb Claude Code Power Users (TechCrunch)
Gemini Apps Usage Limits (Google Help)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.