OpenAI doubles Work and Codex allowances across all three Pro tiers, cancels the Pro 200 cut, keeps Ultrafast exclusive to the $500 plan and reopens Pro 200 to new subscribers

Where this comes from. Both primary materials are OpenAI's own: the help centre article About ChatGPT Pro tiers, showing Updated: 5 hours ago when fetched and already containing the new allowance language, and the announcement posted on X on October 29 by the product lead known as Tibo, handle @thsottiaux. The parts about taking effect that day and about Dots usage staying separate come from that X post as quoted by outlets including BleepingComputer; the parts about tiers, prices and which plan gets Ultrafast come from the help centre.

Start with what was cancelled, because it defines the nature of the announcement. According to BleepingComputer, OpenAI had planned to reduce the included allowance for existing Pro 200 subscribers after October 29 while keeping the monthly price at $200. That plan came from OpenAI's own September 29 email to users. So this is not a promotion. It is a reversal: the company notified a cut, then publicly withdrew it and added capacity on top.

Tibo's wording is the most direct primary account: we are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today; we heard your kind feedback and will not reduce Pro 200 limits, so Work and Codex usage stays the same; extended limits for dots continue and dot usage remains separate from default usage. Three checkable facts sit in that post: the scope includes Business Premium, the effective date is the same day, and Dots is counted separately from the new allowance.

The help centre states the doubling more precisely, and the precision is less flattering. The text says Pro 100 and Pro 500 now include twice their previous Work and Codex allowance, while on Pro 200 only the lower allowance doubled. That means two Pro 200 subscribers paying the same $200 had been receiving different allowances, and this change brings the lower one up to match the higher one rather than multiplying everyone's existing base by two. If you are an existing Pro 200 user already on the higher allowance, your allowance did not move.

The three tiers now line up like this: Pro 100 at $100 a month without Ultrafast, Pro 200 at $200 a month without Ultrafast, and Pro 500 at $500 a month with Ultrafast included. The help centre table lists all three side by side and adds that Pro 200 includes more usage than Pro 100 while Pro 500 offers the highest included usage of the three. It also answers an easy misunderstanding directly: the updated Pro 200 allowance does not include Ultrafast, which is available only on Pro 500.

Ultrafast deserves its own paragraph. The help centre opens by saying Pro 500 includes GPT-6 Astra Ultrafast and GPT-6.1 Sol Ultrafast, meaning the fast variants of those two models are available only on the top tier. That is a harder constraint than a quota: quotas can be traded for time, since waiting longer or running more attempts eventually gets the work done, whereas a reasoning tier gated behind $500 is a structural difference for latency-sensitive workflows, not something thrift can buy around.

How existing users get the new allowance is stated efficiently: new subscriptions include the updated allowance, and existing subscribers receive it automatically without waiting for renewal. That closes a common trap, since many people assume changes land at the next billing cycle and wait two weeks for nothing. The same section answers three edge cases: cancelling and resubscribing yields the same allowance as other Pro 200 subscribers; if the subscription is active you do not need to cancel and resubscribe; and a scheduled cancellation does not end access immediately, so it runs to the end of the current billing period and can be undone before then.

Business Premium appears only in Tibo's post and in secondary reporting; the help centre article on Pro tiers does not mention it at all. The gap is worth flagging: the official documentation covers the three consumer Pro tiers, while the Business Premium doubling currently rests on an X post and media relay. The point is not to nitpick but to warn anyone making a team purchasing decision against copying consumer-tier rules straight onto an enterprise tier.

Why Work and Codex are named together. BleepingComputer states that the change primarily affects ChatGPT Work and Codex, which share usage allowances under eligible subscriptions. That explains a trap many people have hit: you think you only chatted a bit more, but what got consumed was the quota you were saving for Codex runs. For those users the doubling is worth more than it looks, because it relieves the squeeze between two very different workloads sharing one pool.

Put it in the competitive frame and the speed of the retreat makes sense. In the same week Anthropic publicly filed its S-1, Gemini 4 Carbon took the top score of 59 on the Artificial Analysis intelligence index, OpenAI itself shipped the cheaper GPT-6.1 Terra on October 26 and rewrote its Model Spec on October 29. At that tempo a subscription tier is not just a revenue line but a habit hook determining which workbench people stay in, and the opportunity cost of cutting allowances now clearly outweighed the compute saved.

Boundaries. First, the help centre gives no concrete numbers for Work or Codex allowances and says nothing about what happens when they run out; the official language is relative, twice and levelled. Second, the Business Premium scope and figures rest on the X post and media coverage because the documentation does not cover them. Third, Ultrafast refers to the fast variants of GPT-6 Astra and GPT-6.1 Sol; this article does not assess either model's capability or pricing. Fourth, it does not judge whether the allowances are sufficient, because sufficiency depends on your workload and OpenAI publishes no convertible token figure.