On October 29, 2026, OpenAI announced an expansion of ChatGPT Pro usage limits. On the $100-per-month "Pro 100" and $500-per-month "Pro 500" plans, the usage allowance for ChatGPT Work and Codex is now twice what it was. On the $200-per-month "Pro 200" plan, subscriptions that already had the higher allowance keep it, while those that had the lower allowance are doubled, so all Pro 200 subscriptions now sit at the same level.
The key point of this change is that the Work/Codex allowance is managed separately from the message limit for using GPT-6 Pro in regular Chat. Work is an AI agent feature for tasks such as research and document creation, and Codex is an AI agent for software development. The two share the working allowance included in a plan. Using GPT-6 Pro in Chat, on the other hand, has its own weekly message limit, so the amount available for every feature has not simply doubled across the board.
In addition, in Work and Codex, how quickly your allowance is consumed depends on the model, reasoning settings, and processing speed you choose. For example, Fast mode, which speeds up processing, consumes the included allowance 2.5 times as quickly as Standard. So even with a doubled allowance, changing the speed setting can reduce the amount of work you can actually get done. This update is not just about having more to use; it also raises the question of whether to spend your allowance on volume of work or on speed.
Pro 200 is not a flat doubling but a unification at the previous higher allowance
The treatment of Pro 200 is the part that most needs attention. On Pro 100 and Pro 500, the Work/Codex allowance has doubled, but on Pro 200 the change depends on the subscription. Users who previously had the higher allowance keep that level, and only those who had the lower allowance were doubled. In effect, every Pro 200 subscription is now unified at the previous higher level.
OpenAI's official help article, "About ChatGPT Pro tiers", explains this change by separating the Work/Codex allowance from the Chat allowance. The changes to each plan are summarized below.
| Plan / pre-update subscription status | Monthly price (USD) | Work/Codex allowance | Weekly GPT-6 Pro message limit in Chat |
|---|---|---|---|
| Pro 100 | 100 | 2x previous | No change |
| Pro 200 (previously higher allowance) | 200 | Higher allowance maintained | Lower weekly limit unified to the higher one; previous higher limit maintained |
| Pro 200 (previously lower allowance) | 200 | Doubled, unified at the higher allowance | Lower weekly limit unified to the higher one; previous higher limit maintained |
| Pro 500 | 500 | 2x previous | No change |
| Plus | — | No change in this update | GPT-6 Pro not included |
Note: The comparison is before and after the October 29, 2026 update. The Pro 200 Work/Codex allowance and the Chat weekly message limit are separate limits, and a subscription that had a higher allowance for one before the update did not necessarily have a higher allowance for the other.
The weekly GPT-6 Pro message limit in Pro 200's Chat is also being unified at the previous higher level. However, OpenAI has not disclosed specific message counts or the multiplier of the increase. The announcement also does not give absolute allowances for each plan, so it is not possible to calculate from this update alone how many times larger Pro 500's allowance is than Pro 100's.
The ChatGPT release notes dated October 29 also state that OpenAI has withdrawn the reduction in usage limits it announced to Pro 200 users in a September 29 email. OpenAI's Thibault Sottiaux (Tibo) also explained in a post on X the same day that the reduction was called off following user feedback. Users who had the higher allowance will therefore not see the planned cut.
The new allowances are applied automatically to existing subscriptions, with no need to wait for the next renewal date. There is no need to cancel and resubscribe to receive the change, and the updated allowances also apply to new subscriptions. Monthly prices remain $100 for Pro 100, $200 for Pro 200, and $500 for Pro 500; the announcement includes no price reductions.
The expansion applies not only to the individual Pro plans but also to Business Premium. Sottiaux explained that Business Premium's allowance will also double, but the general Business plan as a whole is not covered by the same update. Plus allowances were also not changed in this announcement.
Chat's weekly message limit and the shared Work/Codex allowance are managed separately
ChatGPT Work and Codex share the working allowance included in the same plan. For example, if you use up allowance running large-scale research or document creation in Work, the amount left for software development in Codex decreases accordingly. OpenAI's pricing and usage documentation likewise explains that usage in both is deducted from a common pool. Work and Codex are not each given an independent allowance of the same size.
When you use GPT-6 Pro in regular Chat, on the other hand, a separate weekly message limit applies. In this update, the weekly GPT-6 Pro limit in Chat was left unchanged for Pro 100 and Pro 500, and only Pro 200 was unified at the previous higher level. A larger Work/Codex allowance therefore does not mean the number of times you can use GPT-6 Pro in Chat has grown by the same proportion.
The Work/Codex allowance is also not a guarantee of a fixed number of requests. According to OpenAI's help article on managing usage, consumption varies with the model, the nature of the task, reasoning settings, and other factors. Consumption can rise not only with long inputs and outputs but also when conversation history has accumulated or when tools such as search are used heavily. Because the processing required also changes with factors such as cache usage, you cannot accurately predict consumption from prompt length alone.
For example, a request to fix a few lines of code and a request to investigate multiple files and make large-scale changes can require very different amounts of computation even if each is a single instruction. Long-running tasks executed in the cloud may also consume more allowance than short requests completed locally. The change should therefore not be read as "twice the allowance means twice as many requests can be handled."
The time windows over which limits apply also deserve attention. According to OpenAI's official explanation, Pro 100, Pro 200, and Pro 500 currently have no 5-hour usage limit for Work/Codex. However, the usage included in the plan is itself finite, and separate limits such as weekly limits may apply. The absence of a 5-hour limit does not mean unlimited use of Work/Codex.
The always-on personal AI agent "Dots" is also treated differently from Work/Codex. In his October 29 post on X, Sottiaux explained that the extended allowance set up for Dots will continue for the time being and that its usage will be managed separately from the regular allowance. In other words, usage consumed in Dots is not simply deducted from the shared Work/Codex pool. That said, this was not an announcement that the allowance doubling applies to Dots as well, nor a promise that the extended allowance will continue permanently.
Even with a doubled allowance, choosing Fast can reduce how much you can process
One reason a larger Work/Codex allowance does not necessarily let you process the same amount of work is the difference in consumption rates by speed setting. OpenAI offers settings that change the model's processing speed, such as Standard, Fast, and Ultrafast. These affect not only response speed but also how quickly the allowance included in the flat-rate plan is consumed.
According to the official speed settings documentation, Fast mode consumes the included allowance for supported models at 2.5 times the Standard rate. Ultrafast, available with GPT-6 Astra and GPT-6.1 Sol, consumes at 8 times the Standard rate. These are consumption multipliers for the allowance only; they do not mean that Fast processes 2.5 times faster or that Ultrafast is necessarily 8 times faster.
| Speed setting | Included-allowance consumption rate relative to Standard |
|---|---|
| Standard | 1x |
| Fast | 2.5x |
| Ultrafast (GPT-6 Astra / GPT-6.1 Sol) | 8x |
There are also plan limits on Ultrafast. Among the individual Pro plans, only the $500-per-month Pro 500 includes Ultrafast. On Pro 100 and Pro 200, Ultrafast is unavailable even if you buy additional credits. This means not just that the amount of available credit differs, but that the speed settings you can choose differ by plan.
Let us consider how speed settings change the amount you can actually process on Pro 100 and Pro 500, whose allowance doubled this time. Take a user who used Standard before the update and switches to Fast afterward. To keep the comparison simple, assume the same model, the same input and output scale, and the same reasoning settings, with the usage required for processing unchanged.
If the amount you could process on Standard before the update is 1, then on Standard after the update the allowance has doubled, so you can theoretically process twice as much of the same work. Fast, however, consumes the allowance at 2.5 times the Standard rate, so the amount you can process on Fast after the update is as follows.
\[ \frac{2}{2.5}=0.8 \]
In other words, even with a doubled allowance, switching from Standard to Fast means that under the same conditions you can process the equivalent of 80% of the pre-update Standard amount. This is the result of combining the allowance expansion rate with the speed setting's consumption multiplier.
| Comparison condition | Processable amount, with pre-update Standard as 1 |
|---|---|
| Before update: Standard | 1.0 |
| After update: Standard | 2.0 |
| After update: Fast | 0.8 |
However, this calculation is a theoretical comparison for processing the same amount of work with the same model; it does not measure the number of tasks actually completed, processing time, or output quality. If you keep using Fast both before and after the update, the allowance itself doubles, so the amount you can process under the same conditions also theoretically doubles. For Pro 200 subscriptions that keep the previous higher allowance, the allowance has not doubled in this update, so this calculation cannot be applied as is.
How to make use of this change depends on what users prioritize. Staying on Standard lets you put the larger allowance toward more work. Switching to Fast consumes more allowance but may shorten waiting time. The expanded allowance widens the choice between increasing the volume of work and prioritizing processing speed.
Note that the multipliers above apply when consuming the allowance included in the flat-rate plan. When using purchased additional credits, the consumption rate is 2 times Standard for Fast and 6 times for Ultrafast. These differ from the 2.5x and 8x for the included allowance, so you need to use the corresponding multipliers when comparing allowances and when calculating additional-credit costs.
Compare real capacity by output of the same quality
In evaluating this expansion, differences in consumption by model and reasoning settings also need to be considered. In general, settings that reason more deeply increase the computation required and can consume more allowance. But as OpenAI also explains in its official help, higher reasoning settings do not always produce better results. For some tasks, lighter settings reach sufficient quality, while complex problems may require more advanced reasoning.
Comparing only how fast the allowance drops therefore cannot tell you which setting is more efficient. Even if consumption is low, repeated revisions because the output does not reach the needed quality may end up costing more overall. Conversely, always using the heaviest model and reasoning settings is not necessarily sensible for every task.
To judge real cost-effectiveness, it is useful to compare how much allowance it takes to finish the same type of work to the required quality. By holding the model, reasoning settings, speed mode, and amount of input material constant and checking the remaining allowance before and after, you can more easily see how much work each setting can handle. In ChatGPT, you can check your remaining allowance and reset time by opening "Usage" from "Settings."
Note also that switching models does not restore your allowance once you hit the limit. Work and Codex consume a common allowance, so even if you change to a different model after exhausting the allowance on one, the shared allowance itself is not restored. Changing models can adjust the pace at which you consume what remains, but it cannot recover allowance already used.
After using up your allowance, you can also continue working by buying additional credits for the relevant feature. According to OpenAI's official help on additional credits, eligible features first consume the allowance included in the plan, and then usage is deducted from purchased credits. Eligible features include Codex and ChatGPT Work, but additional credits do not lift the limits on every model or feature in Chat.
These additional credits should also be distinguished from OpenAI API usage fees. When Codex is used with a ChatGPT account, the plan's shared allowance and additional credits apply. When Codex is used with an API key, token-based API billing applies, and the consumption multipliers set for the ChatGPT included allowance are not applied as is. Because the pricing structure differs by how you use it, costs cannot be compared directly even when the same model is used.
With this update, the Work/Codex allowance doubles on Pro 100 and Pro 500, and on Pro 200 the differences in allowance between subscriptions are eliminated. For Pro 200 users who had kept the previous higher allowance in particular, the withdrawal of the planned reduction is significant. At the same time, the working allowance and Chat's weekly message limit are managed separately, and consumption rates also change with speed settings, so simple multipliers alone cannot tell you how usable the plans actually are.
From here, what matters for users is how to use the expanded allowance. Comparing consumption and time required to reach output of the same quality makes it easier to decide whether to stay on Standard and increase the volume of work, or to use faster settings such as Fast. This change expands the computing resources available, but whether you can get the most value from it depends on choosing the model and speed setting that fit the work.