ChatGPT Pro 200 Grandfathering Ends On 29 October: What The Lower Allowance Means For Codex Budgets And Reimbursed Seats

OpenAI now sells three ChatGPT Pro plans. Pro 100 costs 100 US dollars a month, Pro 200 costs 200 and Pro 500 costs 500, and only Pro 500 includes the Ultrafast modes for GPT-6 Astra and GPT-6.1 Sol. Pro 200 was reopened to new subscribers after a pause in September, with a lower included usage allowance than the plan used to carry. Existing subscribers are protected for a short time. According to the OpenAI Help Center, anyone who held an active Pro 200 subscription at any point from 22 September 2026 to 10 a.m. Pacific Time on 29 September 2026 can keep the previous allowance through 29 October 2026, as long as the subscription stays active. After that the subscription moves to the lower allowance and the price stays the same. That date is next Wednesday.

OpenAI does not publish the old and new allowances as fixed numbers. Its Help Center says only that Pro 200 includes more usage than Pro 100 and that Pro 500 has the highest included usage of the three. The ratios that circulated in September came from posts by Tibo Sottiaux of OpenAI, who described Plus as the 1x baseline, Pro 100 as 5x and the new Pro 200 as 10x, down from 20x. Treat those figures as indicative rather than contractual. The safe working assumption is that a heavy Codex user on a grandfathered Pro 200 seat will reach the weekly limit noticeably sooner from 30 October, and will either stop, buy credits or ask for an upgrade to Pro 500.

The detail that matters most for cost is how the allowance is consumed. For the same model, Fast mode uses the included allowance at 2.5 times the Standard rate, and Ultrafast on Pro 500 uses it at 8 times the rate before drawing on credits at 6 times the Standard credit rate. Credits can be bought for Codex, ChatGPT Work and the Word, Excel and PowerPoint add ins, and they are spent after the included allowance runs out. In practice, a developer who leaves Fast mode switched on can use up a smaller allowance in a couple of days and then move to metered spend without anyone noticing. This is the same pattern we described for the Cursor and Grok bundle earlier this week: the sticker price stays the same, and the real cost moves into usage that is harder to see.

For most companies the bigger issue is who owns these seats. ChatGPT Pro is a personal plan. When engineers expense a Pro 200 subscription to use Codex on company repositories, the organisation is paying for a tool it cannot administer. There is no workspace admin console, no SSO enforcement, no organisation owned audit log and no central view of which model touched which code. The allowance change is a useful prompt to find those seats, because the people affected will soon ask for a bigger plan, more credits or a reimbursement for Pro 500. Before approving any of that, check expense claims and corporate card statements for OpenAI charges, and list every individually billed ChatGPT plan that is used for work.

The supported route for company use is a ChatGPT Business or Enterprise workspace, where admins control connectors, retention and access, and where usage can be reviewed. OpenAI release notes from 29 September say dots will not draw on eligible Pro, Business and Enterprise allowances for the first month after launch, which shows that plan terms are still moving between products. If your team depends on Codex, compare the cost of moving heavy users to a managed workspace with paying for Pro 500 plus credits on personal accounts. Include the control gap in that comparison, not just the dollars. A cheaper personal plan that sits outside SSO and offboarding is not cheaper once an auditor asks who had access to production code.

There is a compliance angle as well. Under ISO 42001 a change to the capacity or default behaviour of an AI system in use should be recorded and assessed, and a forced move to a smaller allowance with a faster pay as you go fallback is that kind of change. Record it in your AI system inventory with the effective date of 30 October 2026, note which models and modes are approved, and say whether Ultrafast and purchased credits are allowed. For SOC 2, the relevant controls are CC6.1 on logical access and CC9.2 on vendor risk. If personal Pro seats touch customer data or production code, the vendor record for OpenAI in Vanta, Drata, Secureframe, Sprinto, Thoropass or Hyperproof should say so plainly, with a review date and a named owner, rather than leaving the seats undocumented.

For teams that decide to keep personal Pro plans for now, a few measured checks reduce the surprise. Ask each Pro 200 user to export the usage view from Settings, Usage in ChatGPT, or Usage and Billing in Codex, and to note how much of the weekly allowance they used over the last four weeks. Compare that against the lower allowance once it applies, using the first full week after 30 October as the baseline rather than an estimate. Set a written default that Standard mode is used unless a task needs Fast mode, and that nobody buys credits on a reimbursed plan without approval. These are small steps, but they turn a vague pricing change into numbers that finance can act on.

A short plan for this week. Find every ChatGPT Pro subscription paid for or reimbursed by the company. Confirm with each holder whether they qualify for grandfathering, because the Help Center says only affected users receive an email. Decide before 29 October whether heavy Codex users move to Business or Enterprise, stay on Pro 200 at the lower allowance, or move to Pro 500, and record the decision and its owner. Update your AI acceptable use policy to cover Fast mode, Ultrafast and credit purchases on any plan the company pays for. Our policy templates include an AI acceptable use policy you can extend, and our Codex, Claude Code and Cursor comparisons cover where each tool now sits on price and data handling.