ChatGPT's $200 Pro plan halves its allowance on 30 October and becomes two $100 plans in one

From 30 October, the $200-a-month ChatGPT Pro plan gives you 10 times the Plus allowance in Work and Codex instead of 20 times, for the same price. People who already held the plan in late September keep the old allowance until 29 October; anyone who signed up since has been on the new one from day one.

Until last month the plan was called “Pro 20X”. On 31 August Thibault Sottiaux, who runs Codex at OpenAI, told a user on X: “The Pro 20X is quite precisely 20X the usage of the Plus subscription, so it does exactly what it says on the tin.” On 29 September, the day of DevDay, he posted a new scale: “Plus = 1X, Pro 100 = 5X, Pro 200 = 10X.” The multiple left the name and the price took its place. One Hacker News reader searched the upgrade screens and found only “Pro 100”, “Pro 200” and “Pro 500”, with no 10x, 20x or 25x anywhere. OpenAI’s help centre says only that new subscriptions include “a lower usage allowance than previously offered with Pro 200”.

So the 10x comes from an OpenAI employee speaking in public. Two other numbers, GPT-6 Pro in Chat falling from 200 to 100 messages a week and the new $500 tier being worth 25 times Plus, so far appear only in OpenAI emails that subscribers have posted online. The passage one subscriber pasted on Hacker News on 30 September reads: “Starting October 30, 2026, your included usage in ChatGPT Work and Codex will decrease from 20 times to 10 times the ChatGPT Plus allowance. Your GPT-6 Pro limit in Chat will decrease from 200 to 100 messages per week. Your price remains unchanged and you’ll keep your current limits through October 29, 2026.” OpenAI’s help article on GPT-5.6 and GPT-6 Pro lists 200 GPT-6 Pro messages a week for Pro 200, which matches the email’s starting point.

A Plus-sized block of usage now costs $20, not $10

On the multiples, the arithmetic is blunt. $200 used to buy 20 Plus-sized blocks of usage at $10 each, half the price of Plus itself. From 30 October each block costs $20: the same as Plus, the same as Pro 100 ($100 for 5x) and, on the email’s 25x, the same as Pro 500. The Pro plans used to be a bulk discount. Now Pro 200 is two Pro 100s in one box, and getting back to 20x or more means Pro 500 at two and a half times the price.

Chat works out the same way. The help centre gives Pro 100 a single pool of 50 messages a week shared between GPT-6 Pro and GPT-5.6 Sol Pro. Over a month of 4.33 weeks, a GPT-6 Pro answer on Pro 200 cost about $0.23 before the cut and about $0.46 after it, exactly the Pro 100 rate.

Plan	Per month	Agent allowance (Work and Codex)	Notes
ChatGPT Plus	$20	Baseline	Five-hour windows plus a weekly cap
ChatGPT Pro 100	$100	5x Plus	50 GPT-6 Pro / 5.6 Sol Pro messages a week, shared
ChatGPT Pro 200, September	$200	20x Plus	200 GPT-6 Pro messages a week
ChatGPT Pro 200, from 30 October	$200	10x Plus	100 GPT-6 Pro messages a week (email)
ChatGPT Pro 500	$500	25x Plus (email)	Only personal plan with Ultrafast
Claude Max 5x / 20x	$100 / $200	5x / 20x Claude Pro per five-hour session	Separate weekly cap; $100 / $200 a month in API credits
Doubao Professional	¥500 (about $75)	10x the ¥68 Standard plan (per Huxiu)	Five-hour allowance plus a weekly cap

Yuan converted at the ECB reference rate of 23 October, $1 = ¥6.6702. ChatGPT and Claude are not offered in mainland China; Doubao is ByteDance’s assistant.

One more rule on OpenAI’s Codex pricing page is easy to miss: Fast mode draws on your allowance at 2.5 times the Standard rate. If you leave Fast on, the new Pro 200 is worth about 4x Plus at Standard speed.

Who OpenAI’s argument works for

In his 29 September post Sottiaux conceded that “if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan”. His case is that models keep getting cheaper and more efficient: that week GPT-6 Sol and GPT-6 Luna launched at half their previous API prices, so subscribers “will still get more work done than if you were on the Pro $200 subscription one month ago”. In exchange, OpenAI committed not to bring back the five-hour limit on Pro, so a week’s allowance can be spent in a burst.

Aris, lead writer at the hardware review site Hardware Busters, is a Pro 200 subscriber whose lab uses Codex every day for test applications and report tooling. He read the email “the way any business reads a supplier doubling its prices”, and wrote: “A smarter model is the product… That has never been a reason to halve what a customer already pays for.”

OpenAI’s own credit rate card shows where the line falls. GPT-6.1 Sol costs 250 credits per million output tokens and the flagship GPT-6 Astra 1,250, a factor of five. If you can move routine work from Astra to 6.1 Sol, half the allowance may well get more done than September’s did. If you depend on Astra, or mainly use GPT-6 Pro in Chat, which runs on Astra and simply has half the messages, you get half.

At the same $200, Claude Max 20x looks like twice as much, but Anthropic’s help centre says its 20x is 20 times Claude Pro’s “per-session usage allowance”, with sessions resetting every five hours and a separate weekly limit whose size it does not publish. Sottiaux’s “exactly what it says on the tin” was a reply to a user complaining about precisely that: that Claude’s 20x covers only the five-hour window and that its weekly limit is “basically just 2x the $100 plan”. ByteDance’s Doubao, at ¥500 (about $75) for its top tier, also sells only multiples. Each company multiplies an unpublished base, so the multiples cannot be compared across companies.

OpenAI’s base is unpublished too. 10x means ten times Plus’s weekly allowance, and OpenAI has never said what Plus gets in a week. The Codex pricing page estimates only what Plus gets per five hours, 15 to 160 local GPT-6.1 Sol messages, a tenfold range, followed by “weekly limits may also apply”. Subscribers see a percentage in Settings. On 11 October a self-described long-standing Pro 200 subscriber posted on OpenAI’s developer forum that his allowance was draining much faster after a reset and that he feared the cut had come early, but “I cannot verify that from the percentage meter alone”. His post is still the only one in the thread.

Decide by your heaviest week

Existing subscribers have a cushion. Sottiaux said they would “receive a lot of additional credits”; according to the subscriber emails that have been published, it is 62,500 credits, valued by OpenAI at $2,500 and expiring on 31 December. The US App Store sells 100 credits for $4, which matches. Credits are spent only after the included allowance runs out, so they can absorb overflow in November and December, and there is no need to jump to Pro 500 now. The decision can wait until January.

Before 29 October it is worth opening Settings, then Usage, to see what share of the old allowance your heaviest week used. The new Pro 200 is half the old allowance and Pro 100 a quarter. If even your heaviest week stayed under a quarter, Pro 100 saves $100 a month and still fits. If you used more than half and work mostly in Codex, moving what you can to GPT-6.1 Sol is cheaper than jumping to Pro 500. People who rely on GPT-6 Pro in Chat have no such way out: switching models does not help them, and Pro 100 offers only a shared 50 messages a week.