OpenAI Canceled the Pro 200 Cut and Doubled Codex Limits. Don't Build Your Workflow on a Quota That Changes by Press Release.
OpenAI canceled a planned cut to the Work and Codex allowance on its $200 Pro plan and doubled the limits on its Pro tiers, effective October 29. The cut would have taken Pro 200 from roughly 20 times the Plus allowance to 10 times, a 50% reduction by those multipliers. If you code with Codex all day, you got a good week.
I would still not relax. The reversal followed subscriber backlash, which means the quota was one announcement away from halving, and your workflow was priced against it the whole time.
What changed
Per Kingy AI's rundown, which cites OpenAI's updated Pro guide and its October 29 release notes, the Work and Codex allowance doubled on Pro 100 and Pro 500. Pro 200 accounts on the lower allowance were raised to the higher one, and accounts already on the higher allowance keep it. Plus is unchanged. List prices are $100, $200 and $500 a month for the three Pro tiers.
Business Premium usage also went up, but I could not find a published multiplier, so I will not guess one. Premium seats are $125 per user per month billed monthly, or $100 on annual billing, with a two-seat minimum.
One caveat on the numbers: the 20x and 10x multipliers come from earlier reporting, not from a newly published quota table. OpenAI does not give you a fixed token count, so every "double" is relative to a number you cannot see.
The real cost is the planning you couldn't do
Subscription quota is a discount, not infrastructure. You pay a flat fee, and in exchange the provider decides how much work that fee buys, and can change its mind at any time. That is a fine deal for a person. For a business it is a supplier who can change your unit cost without a contract.
Think about what the planned cut would have done to a solo operator running an agent loop on Pro 200. The same fee, half the work. If you had quoted a client or set a launch date assuming the old allowance, you would have found out from a forum thread.
I do not say this to pick on OpenAI. Every flat-rate AI plan has this shape, and every provider has tightened limits on some tier at some point. The point is that a flat fee hides a variable cost, and the variable cost shows up the moment the vendor changes the rules.
What I'd actually do
Price your work at the API rate, even if you pay the subscription. Take a typical week, estimate the tokens or task count, and multiply by what the same work would cost through the API. That number is your real cost of goods. The gap between it and your subscription is margin you are borrowing from the provider, and you should know how much.
Then keep a tested fallback. If Codex limits halve on a Tuesday, I want to be able to point the same job at another coding agent or an API key within an hour, not discover that my prompts only work in one tool. And do not prepay annual plans on the strength of today's limits. A monthly plan is cheap insurance against a policy you can't see coming.
The honest counter-take
There is a fair argument that this episode shows the system working. OpenAI proposed a cut, users objected loudly, and the company reversed course and went further than it started. Companies that listen are better than companies that don't. And for heavy Codex users, a flat $200 is far cheaper than API billing would be, so the subscription is still the right buy for most individuals.
I agree on both counts. My claim is narrower: buy the subscription, but do not build a business plan on it. Treat the doubled limits as a gift, enjoy them, and keep the spreadsheet that tells you what you would pay if the gift ended next month.
Author
Lukas
@lukcombinator