ChatGPT Pro 500 vs Pro 200: keep the $200, move to $500, or cancel
If half of last week's Codex and Work still finishes, keep ChatGPT Pro at $200. Move to the $500 plan only when you already hit the wall. If you only chat, cancel.
The repo is still open. The afternoon change has not finished its tests, and you open Settings, then Usage, to see what is left on the Codex and Work bar. The email says the old allowance lasts through October 29. The pricing page has a new $500 month next to the $200 one. You want the bill: renew $200, switch to $500, or cancel this month.
People are already searching Should I use ChatGPT Pro, Is ChatGPT Pro worth $200, ChatGPT Pro 500 vs Pro 200, and Is Ultrafast worth $500. Same person. A different week of work. The opposite answer.
Keep paying $200
If the bar is still above half on Sunday night, this cut does not reach you.
On September 29, 2026, Codex lead Thibault Sottiaux (Tibo) wrote the multipliers as Plus = 1X, Pro 100 = 5X, and Pro 200 = 10X. He also said that, in API list-price dollars, the new $200 plan comes out to about half the API spend of the old Pro 200. People already subscribed keep the 20X multiplier for a while, and he said they receive additional credits. The Help Center page on Pro tiers does not print "10X" in a table. It says new Pro 200 subscriptions that are not eligible to keep the old allowance include a lower allowance than Pro 200 used to, "to reflect our increasingly efficient models," and the monthly price stays $200. The percent by which token consumption fell is not on that page. That percentage is missing. Leave it off the invoice.
Who keeps the old allowance through October 29 is fixed there. If your Pro 200 was active at any point from September 22, 2026 through 10 a.m. Pacific Time on September 29, 2026, you can use the previous included allowance through October 29, 2026, while that Pro 200 subscription stays active. After that date the allowance drops. The price stays $200 a month. Keeping the old allowance does not move you to Pro 500, and it does not add Ultrafast.
Inside that window, if the subscription ends and you subscribe again on the same account, the old allowance still only lasts through October 29. The days you were unsubscribed are not saved for later. Outside that window, a subscription that has ended and is bought again uses the lower allowance. Scheduling a cancellation does not cut you off immediately. You keep the plan until the current billing period ends. The amount of those extra credits, and the date they expire, are missing from the public pages. With no number, leave them out of this month.
The line for staying is one test. Take the Codex and Work you actually finished last week and cut that week in half. If the halved week still finishes before the reset, stay on $200. Tests go red and you rerun them. The usage bar does not go red first. The bill is still $200, and you are not saving a half-written diff as a draft until next week's reset. Run the same test after October 29. If the new allowance still covers your week, $200 is still your plan. If it does not, that is the next section. Do not add the $500 plan in this section just in case.
Do not press $500 for the speed switch. The same Help Center page says that, for the same model, Fast consumes included allowance at 2.5 times the Standard rate, and GPT-6 Astra Ultrafast consumes it at 8 times. Those are allowance rates, not a promise the answer arrives faster. If you would not turn that switch on, $500 is a switch you will leave off. How many times Plus the Pro 500 allowance is, the Help Center and those two posts do not say. That multiplier is missing. Do not divide a missing number into the $200 plan.
Tibo also said the $200 plan will not get the 5-hour limit back, so you can spend the weekly allowance on your own schedule. The Pro tiers page does not say that. The Work and Codex usage article still says some plans have both a five-hour window and a weekly window, and the allowance does not correspond to a fixed number of messages or tasks. Read the reset time on your own Usage screen. The post does not replace that line. His multipliers are in the posts mirrored at the 10X note and the half-API-spend note. Those pages are mirrors, not the original x.com posts.
Move to the $500 plan
If you were already empty on Wednesday or Thursday under the old 20X, with the change sitting in the repo and the same diff pasted back in the next morning, the extra $300 starts to look like a bill. The terminal stops mid-run. You paste the prompt again. The model rereads files it already read yesterday. You wait on the same error until the reset. If that afternoon is most weeks, move to Pro 500. If it happened once this month, wait for the reset, or buy credits after the included allowance is gone and finish that one job. That is usually cheaper than the upgrade. Pro is billed monthly. A one-month trial is a $500 charge. The release notes say Pro 500 is $500 a month, has the highest included usage of the Pro plans, and is the only Pro plan with Astra Ultrafast in ChatGPT Work and Codex. They do not say that usage is high enough for your repo. Whether it covers you is still the day you hit the wall last week.
You move up only when both of these are true for your week. After the drop to 10X, the repo does not finish in a week, so you need the higher included usage. You also need Ultrafast, and only on the runs where you are stuck. Engadget on September 29, 2026 wrote that with Ultrafast on, Codex and Work generate up to 300 tokens per second. The Help Center does not state tokens per second. It states the 8x consumption rate. If you leave Ultrafast on as the default, including for a rename or a comment, you burn the highest allowance faster. Keep the cheaper model as the daily default. Move up to Astra when a wrong change means rework, or when the cheaper pass has already failed. This site already draws that line: when GPT-6 Astra should not be your daily default.
When the bar hits empty, check whether every turn stuffs the whole instruction file into context. The same wasted context costs more at $500. Cut Codex token use before you add money, then decide whether to upgrade.
Credits purchased on Pro 100 or Pro 200 do not unlock Ultrafast. After included usage is gone, credits can pay for eligible use in Codex, ChatGPT Work, and ChatGPT for Word, Excel, and PowerPoint. They do not add usage for every ChatGPT feature. You want the speed tier. Purchases on the $200 plan still cannot turn Ultrafast on.
Leave
You have a chat open, not a repo. You are drafting, rewriting mail, or telling a Dot to watch something. The release notes say that for the next month, dots usage will not count toward the plan allowance for eligible Pro, Business, and Enterprise users. The end date of that month is missing from that paragraph. At launch, Pro access to Dots excludes the European Economic Area, Switzerland, and the UK, and it is for people 18 or older. If you mostly talk to a Dot, the $200 is not this week's Codex allowance. Chat does not hit that bar, so a lower allowance does not hit you either. $500 is not a better chat window. Dots versus Meta Muse is about whether the agent watches the work. It does not make this subscription cheaper.
On that kind of week, leave. The chat window still answers tomorrow. What stops is the $200 charge, not the draft. Drop to Plus, or cancel Plus too if you do not fill it. If you hesitate because the plan is named Pro, check whether the allowance stopped you once last week. If it never did, the name is not worth the charge. After you leave Pro, the $20 choice is ChatGPT Plus vs Claude Pro. That page still describes $200 Pro as the top tier, and the allowance language there is the old one. Use it to pick a $20 tool. Do not use it to decide whether you stay on Pro. If even $20 is too much, use AI subscriptions under $20.
If your Codex week fits in 5X Plus, Tibo's tier for that is Pro 100 at $100 a month. Search already asks about ChatGPT $100 and 10x Codex usage. Pro 100 has no Ultrafast, and its included usage is lower than Pro 200. Staying on $200 means paying for a slice you do not use. How many messages or tokens sit inside Plus at 1X is not published as a fixed number. That figure is missing. Compare it with your own reset record from last week, not someone else's screenshot.
There is a way of leaving that costs you. The allowance still covers the work, and the unchanged price after the cut makes you angry. Anger is not the allowance. Inside the window, the old allowance is still there until October 29. If you cancel to make a point, what you lose is the 20X on the days you are unsubscribed. Keep the anger. Keep the subscription until you watch the new allowance fail to cover a real week.
The drop from 200 GPT-6 Pro chat messages a week to 100 is Engadget's account, and some people say their email says the same. The Help Center article on GPT-6 Pro does not print that new number. It says Pro 200 has a weekly GPT-6 Pro limit and, when you hit it, ChatGPT switches to GPT-5.6 Thinking at Medium. It prints neither 200 nor 100. If you never bottom out GPT-6 Pro in chat, neither figure should push you onto $500 a month.
Look at this tonight
Open Usage. Look at the shared Codex and Work bar.
If it empties before the week's repo is done, it emptied on the old allowance, and you did not leave Ultrafast on by accident, move to $500. Ultrafast eats allowance at 8 times the Standard rate. Turn it on only for the hour you are actually short.
If half the bar still finishes the week, renew $200. Do not compare $500 with a multiplier that is missing.
If the bar has nothing to do with your week because you are in chat and Dots, leave. The next bill is the $20 tier. If 5X is the ceiling your Codex actually hits, take Pro 100 at $100 a month.
Four figures are missing from the public pages checked for this: how many credits you get, the absolute token count behind the new allowance, how many times Plus the Pro 500 allowance is, and the date when Dot usage starts counting. Do not put a missing number on your own bill.
FAQ
Should I use ChatGPT Pro?
Use Pro if Codex and Work are the thing you pay for, and last week's work still finishes after you cut that week in half. If the usage bar never moves because you are in chat, do not use Pro.
Is ChatGPT Pro worth $200?
It is worth $200 when a halved week of your real Codex and Work still finishes before the reset. It is not worth $200 when that bar never stops you. The $200 price stayed. New subscriptions that are not keeping the old allowance have a lower included allowance. Eligible subscriptions keep the previous allowance through October 29, 2026.
ChatGPT Pro 500 vs Pro 200: which bill is yours?
Pro 500, at $500 a month, is yours only if you already ran out on the old 20X in most weeks and you will turn Ultrafast on only when a run is stuck. If half the bar still covers the week, the bill is Pro 200 at $200. How many times Plus is included in Pro 500 is not published.
Is Ultrafast worth $500?
It is worth $500 only if you need that speed on the runs where you are stuck. Among Pro plans, only Pro 500 includes it. For the same model it consumes included allowance at 8 times the Standard rate. Leaving it on for every small edit burns the higher allowance faster. Credits bought on the $100 or $200 plan do not unlock it.
What is 10x Codex usage on the $200 plan, in messages or tokens?
That absolute number is missing. Tibo described the new Pro 200 as 10X Plus and about half the old plan's API-dollar spend. The Help Center does not publish a fixed message count or token count for the allowance. Read Settings, then Usage, on your own account.
Is ChatGPT $100 enough?
Pro 100 is $100 a month. On Tibo's scale that is 5X Plus, with no Ultrafast and less included usage than Pro 200. If your Codex week fits in that 5X, the $200 plan is the slice you are not using. If you already empty a 20X week, $100 is the wrong direction.
Continue exploring
More decisions worth reading
Follow the thread from this article to the next practical buying question.
Plan comparison
01ChatGPT Plus vs Claude Pro (2026): Same $20, Radically Different Tools
Open guidePlan comparison
02Gemini Deep Research vs ChatGPT Deep Research: A Comparative Analysis (2026)
Open guideBuying advice
03One $20 or Two $40? The AI Subscription Stacking Debate (2026)
Open guidePlan comparison
04