Cut One Day, Doubled the Next: OpenAI's Pro Limit Reversal in 24 Hours — Codex Users, Don't Rush to Upgrade

One day you are told your allowance is about to be cut; the next day you are told it will be doubled. That is the 24-hour reversal ChatGPT Pro subscribers experienced on October 29, 2026. The starting point is not complicated: on September 29, OpenAI emailed Pro 200 subscribers to say that, after October 29, the usage allowance included in their subscription would be reduced, while the price would stay at $200 per month. On October 29 itself, OpenAI published no official blog post. Instead, Thibault Sottiaux, the head of ChatGPT and Codex — known on X as Tibo — announced the reversal in a post on X. According to BleepingComputer's report on October 29, 2026, the announcement came down to two sentences: one about doubling limits, and one about Pro 200 not being cut. For people doing real work in Codex, that is obviously good news — but the wording of the good news is far more complicated than the headline suggests.

First, pin down the announcement itself, including its exact wording and its medium, because that determines how much weight every later judgment can carry. BleepingComputer's report was published on October 29, 2026, and it states explicitly that OpenAI issued no official blog post for this change; the vehicle was Tibo's post on X. Tibo's exact words were: "We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today." His other sentence was: "We heard your kind feedback and won't be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same." These two sentences must be read together: the first talks about doubling, the second talks about Pro 200 staying the same. Appearing in the same announcement, they leave an obvious question unanswered: doubling on what base? And how can "doubling" and "Pro 200 stays the same" both be true at once? The report gives no precise per-tier numbers, and OpenAI does not publish exact numerical limits per tier, so that question cannot currently be answered with arithmetic. All that can honestly be recorded is the tension itself.

Now straighten out the timeline. On September 29, OpenAI emailed Pro 200 subscribers that their allowance would be cut after October 29, with the $200-per-month price unchanged. That notice is the origin of the reversal and the source of the backlash: paying the same money for less allowance feels, to any subscriber, like a price increase in disguise. On October 29, Tibo announced the reversal on X, saying it took effect that day. BleepingComputer's report followed later that day, headlined as OpenAI doubling ChatGPT Pro usage limits after backlash over a planned cut. The timeline itself is not in dispute; only the precise meaning of "doubling" is. That distinction must be kept clear: the dates and events are settled, the limit numbers are not, and the certainty of the former must not be lent to the latter.

What does "doubling all limits" actually double? The wording gap must be put on the table

This is the most important section of this article, and also the easiest to get wrong. The X post literally says "all usage limits," which sounds as if everything doubled. But several secondary readings point out that the actual change is much narrower: it mainly affects the shared usage allowance for ChatGPT Work and Codex, which share an allowance under eligible subscriptions; the weekly GPT-6 Pro conversation cap is unchanged, at least for the Pro 100 and Pro 500 tiers; and for Pro 200, the treatment is cancellation of the cut, leveled back to the original higher tier, rather than doubling on top of a reduced base. There is also a formulation circulating in media coverage, describing a move "from 20x Plus down to 10x Plus." This article must label that explicitly: it is a media-report formulation, not a figure confirmed by OpenAI. OpenAI does not publish exact numerical limits per tier, and any article that converts a multiplier into a specific number of uses is filling in numbers the company has not confirmed.

Why spell out the wording gap in such detail? Because for Codex users, different readings lead to completely different actions. If you believe "all limits doubled," you might feel safe handing Codex more work, or think you can downgrade and save money. If you understand it as "only the shared Work and Codex allowance changed, while the GPT-6 Pro conversation cap did not," you will first check which kind of usage you actually rely on most, and only then decide whether to change your subscription. Pro 200 users are in yet another position: for them, the most important news is not doubling, but "no cut." They kept their existing allowance, rather than gaining a new allowance that can be precisely calculated. Collapsing these three situations into one sentence — "everyone got doubled" — is an irresponsible simplification.

One further point needs to be stated separately, to avoid confusion. Extended limits for Dots, the always-on assistants, will continue to be offered, and Dots usage remains separate from default usage. As BleepingComputer relays it, your dot's usage will still be counted separately from your default usage, so it will not eat into your Work and Codex allowance. That is good news for people already using Dots, because it means an always-on assistant will not quietly crowd out the allowance you use for coding. But note the time qualifier in the same statement: Tibo's exact words are "for now" — the extended Dots limits continue for now. That is not a permanent promise, and the qualifier should be read together with the promise.

Two more boundaries should be drawn. First, the Plus tier is unchanged. This announcement covers Pro 100, 200, and 500 and Business Premium; it mentions no change for Plus, and Plus subscribers should not infer from this report that their allowance changed too. Second, there was a separate announcement about API credits on October 28, and it must not be confused with this subscription-limit reversal. That is a different thread and a different product dimension; adding it to the Pro subscription calculation only creates more confusion. This article is about subscription limits only, not API credits.

What is the practical effect for Codex users?

For users who write code, refactor, and run agent tasks in Codex every day, the direct effect of the reversal is this: you no longer need an emergency plan for "my allowance gets cut in half after October 29." Had the original cut gone ahead, heavy users would have been hurt most: long sessions, multi-file refactors, and continuously running agent tasks are exactly the uses that consume the most allowance. Cancelling the cut at least preserves those users' existing working rhythm. And the "doubling" language, if read in the secondary sense as some relaxation of the shared Work and Codex allowance, adds buffer room: you can run somewhat more work without upgrading to a more expensive tier, or face fewer interruptions when an allowance runs out. But how large that buffer really is cannot currently be given as a reliable number, because the company does not publish exact figures. Any article claiming "you can now run this many more tasks" is estimating in a place where the company has confirmed no number.

There is also a practical question of working habits. Codex usage is not consumed evenly; it spikes exactly when you are busiest and closest to a deadline. An allowance that is sufficient on an ordinary week may be completely inadequate in a crunch week; an allowance that looks generous in normal use can drain quickly during continuous agent runs. That is why a change in limit policy affects heavy users far more than light users. Light users may barely notice either a cut or a doubling; heavy users repeatedly test the boundary. That is also why the backlash came mainly from heavy users and Pro 200 subscribers: they were not arguing about an abstract number, but about whether their workflows would be interrupted. Once that is understood, the 24-hour reversal makes sense: OpenAI was not facing a single complaint, but a workflow risk for a core group of users.

For anyone considering an upgrade or a downgrade, this article's advice is very specific: do not move yet; wait for the help pages. An X post is not the same as the numbers on a help page, and there is often a gap in wording and detail between a social-media promise and the formal allowance table in documentation. Before upgrading, confirm at least three things: first, which allowance you actually use most — the shared Work and Codex allowance, the GPT-6 Pro conversation cap, or the separate Dots allowance; second, what the formal wording is on the help page for the tier you want to move to, rather than the summary wording in the X post; and third, what your own current usage curve looks like — whether it occasionally spikes, or runs close to the ceiling all the time. Until those three things are confirmed, any upgrade or downgrade decision is made with incomplete information, and it is easy to regret it at the next policy change.

How to give your own usage a checkup

While waiting for the help pages to be updated, users can already do one thing that requires no official numbers at all: give their own usage a checkup. The method is not complicated. First, recall which days and which tasks consumed the most usage over the past month — whether it clustered on a few crunch days, or stayed fairly even every day. Next, separate where the usage mainly went: Codex agent tasks, Work's everyday conversations and document work, or Dots' always-on assistance. Finally, estimate where your work would have been blocked first if the allowance really had been cut in half: would a long session fail to finish, or would a batch job have to be split across several days? The value of this checkup is that it turns an abstract "allowance" into concrete "workflow breaking points." When the help pages' formal numbers finally appear, you will not need to guess which kind of user you are; you can compare the checkup directly against the new allowance and judge whether it is enough.

The checkup has a side benefit: it can reveal waste that has nothing to do with limit policy. Many users' usage peaks do not come from real workload, but from repeated debugging, repeated calls without caching, or an imprecisely written prompt retried over and over. Cleaning up that waste is equivalent to expanding your own allowance without spending anything. Conversely, if your usage structure is already clean, with almost no waste, then a change in limit policy affects you in full, and upgrade or downgrade decisions should be made more cautiously. Whatever the official numbers turn out to be, understanding your own usage structure first is never the wrong step.

One wording detail is also worth noting. When announcing the reversal, Tibo said "we heard your kind feedback." Readers familiar with this kind of corporate language will know that the subtext is often the opposite: the feedback was not gentle, but intense enough to force a policy reversal. Calling a backlash that caused a reversal "kind" is a classic softening device: it acknowledges users' voices while avoiding describing the situation as a confrontation. Readers need not get hung up on the literal wording; the real information is in the action — the planned cut was cancelled, and the allowance was not reduced. Action is more credible than wording, which is why this article insists throughout on judging by actions and formal documentation, not by social-media phrasing.

Business Premium subscribers deserve a separate note. Tibo's exact words place Business Premium in the same sentence as Pro 100, 200, and 500, saying its usage limits are likewise doubled. But BleepingComputer's report focuses on the Pro tiers and the Work and Codex relationship, and says much less about the specific application to Business Premium. Business Premium users should therefore be especially cautious: do not assume that secondary readings about the Pro tiers can be transferred directly to your own tier. Wait for formal guidance for your own tier before acting. Tiers mentioned in the same sentence do not necessarily share identical implementation details, and that is especially common in subscription policy.

View: the limit itself has become part of the product, and trust costs more than allowance

The question truly worth discussing here is not the number for any one tier, but subscription trust. When one of a subscription product's core selling points is "a generous allowance you can use freely," the limit itself becomes part of the product, not an incidental term. A subscriber paying $200 a month is buying not only access to a model, but also a predictable amount of room: I know roughly how many tasks I can run this month, and I can plan work around that. The September 29 email broke precisely that predictability: same price, less allowance, and users' work plans had to change with it. The October 29 reversal repaired the allowance, but it did not fully repair predictability, because users have now seen that an allowance can be changed on a month's notice.

A deeper problem is the announcement channel itself. Such an important subscription policy change did not go through an official blog post and help-page update first, but was announced by the responsible executive in a post on X. That has a speed advantage — it can respond on the day of the backlash — but it carries a cost in seriousness: an X post is not a formal document, its wording can be general and conversational, and it can leave room for interpretation, while what users really need is precise rules they can act on. When a general phrase like "all usage limits" needs secondary interpretation to narrow it down, the chain of information has already grown longer, and each extra link adds a chance of error. If OpenAI wants this reversal to genuinely settle the dispute, its next step should be a clear, tier-by-tier formal statement on its help pages, not continued reliance on interpretations of an X post.

For the industry as a whole, this is also a signal worth recording. Competition among subscription AI products is expanding from "whose model is stronger" to "whose allowance is more stable." When users choose a Pro tier, they weigh not only model quality, but also whether the allowance is sufficient, whether policy is stable, and whether changes come with enough notice and transition arrangements. A 24-hour reversal looks, in the short term, like OpenAI listening to feedback; in the longer term, it exposes a structural problem: when allowance policy can swing back and forth this quickly, users find it hard to commit long-term workflows to a single tier. Vendors certainly have the right to adjust their commercial strategy, but the frequency and manner of adjustment are themselves part of the product experience. Frequent, large, short-notice adjustments will make users more conservative about upgrading — a loss for both vendors and users.

Finally, the judgment can be condensed into directly usable conclusions. First, for Pro 200 subscribers, the core gain is "no cut," not a doublable number that can be calculated. Second, Pro 100 and Pro 500 subscribers should focus on the change to the shared Work and Codex allowance, not assume that all usage doubled — especially since, on secondary readings, the weekly GPT-6 Pro conversation cap did not change. Third, Dots users can be reassured for now, because extended limits continue and remain separate from default usage, but remember the qualifier "for now." Fourth, Plus subscribers are outside the scope of this change and should not misread it. Fifth, everyone should wait for the formal numbers on the help pages before upgrading or downgrading, and not decide on the basis of a single sentence in an X post. Today's allowance represents only today; it promises neither permanence nor the direction of the next change. Planning with allowance as a variable that can change, rather than relying on it as a fixed asset, is the most practical lesson this reversal leaves for all subscribers. See the rules clearly first, then decide where to spend.