ChatGPT Pro 200 halves on October 30. The cut lands on Astra.

OpenAI's $200 plan drops from 20x to 10x Plus usage on October 30. Sol got 50% cheaper, so its tokens hold. Astra didn't, so Astra users lose half.

The Editors · AI money · 7 min read · Oct 12, 2026

Black laptop computer turned on displaying blue screen

By the numbers

20x → 10xPlus usage in the $200 ChatGPT Pro plan from October 30, 2026

$10 → $20what the $200 plan charges per Plus-sized allowance

$2/$10 vs $10/$50API price per million tokens, GPT-6 Sol (cut 50%) vs GPT-6 Astra (not cut)

8xrate at which Ultrafast, the $500 tier's exclusive, burns the allowance

$2,500credits granted to existing $200 subscribers, gone December 31

On October 30, the $200 ChatGPT Pro plan drops from 20 times the Plus allowance to 10 times, at the same price (Hardware Busters, October 1, 2026). OpenAI's answer is that you'll still get more work done than before. That holds for one model and fails for another.

The allowance is metered in API dollars. OpenAI's Thibault Sottiaux said the new plan "nets out at half the dollar in API spend" of the old one (explainx, September 29, 2026). On September 22, OpenAI halved the API price of GPT-6 Sol, from $4 to $2 per million input tokens and from $20 to $10 per million output (The Next Web, September 22, 2026). Half the dollars at half the price buys about the same Sol tokens.

GPT-6 Astra, the flagship, still lists at $10 and $50 (CloudZero, September 4, 2026). If most of your Pro usage runs on Astra, you lose about half of it on October 30. If it runs on Sol, the cut is mostly a repricing of the meter.

What changes on October 30

The change applies to ChatGPT Work and Codex alike. Four things move:

Allowance: 20x Plus becomes 10x Plus for $200 a month (Nerdschalk).
GPT-6 Pro chat: 200 messages a week become 100 (Hardware Busters).
Who waits: new subscribers have had the 10x allowance since the plan reopened on September 29. Accounts with an active $200 plan in the week before that keep 20x through October 29, as long as they don't cancel (XenoSpectrum, September 29, 2026).
Compensation: those same accounts got 62,500 usage credits, valued at $2,500, which expire December 31, 2026 (MakeUseOf, September 30, 2026).

OpenAI had paused new $200 sign-ups on September 10 because demand for Astra outran capacity (Christopher Alarcon, updated October 2, 2026). The plan came back smaller, next to a new $500 tier.

Every tier now costs $20 per Plus allowance

The old $200 plan was the one volume discount in the lineup. It sold 20 Plus allowances for $200, $10 each. From October 30, every paid tier prices the same unit at $20:

Plus: $20 for 1x
Pro 100: $100 for 5x
Pro 200: $200 for 10x
Pro 500: $500 for 25x

The multiples come from Nerdschalk and Christopher Alarcon. One flag: the 25x for Pro 500 is widely reported, and at least one tracker says OpenAI's help page doesn't print it as a number.

So the bigger plans buy headroom, priced the same per unit. If you measured your use at 4x Plus over the last month, Pro 100 covers it, and the extra $100 for Pro 200 buys capacity you won't touch.

Why Astra users take the cut

The meter counts API dollars at list price. That's the claim on record, and OpenAI hasn't published a token table to check it against, so read what follows as arithmetic on that claim.

Take a week of work that costs $100 at API list price under the old Sol rates. After September 22, the same Sol tokens cost $50. Half the dollar allowance still covers them. That's where "you'll get more done" comes from: GPT-6 Sol also needs fewer tokens per task, by Sam Altman's account ("half the price per token, and even less per task", via The Next Web).

Astra had no cut. A week of Astra work that cost $100 at list price still costs $100, and the allowance paying for it is half as big. The explainx write-up flags the same split: Sol users may see no change in tokens, and Astra-heavy users take a real reduction (explainx).

Astra is the model that drove the September sign-up pause. The plan was halved on the meter of the cheap model and lands hardest on the expensive one.

$10$20$20$20$160

View as table

What ChatGPT charges per Plus-sized allowance
label	What ChatGPT charges per Plus-sized allowance
Pro 200, grandfathered to Oct 29	$10
Pro 100	$20
Pro 200 from Oct 30	$20
Pro 500, standard speed	$20
Pro 500, Ultrafast only	$160

Source: Nerdschalk, Christopher Alarcon, Nerdschalk Ultrafast rates, Oct 2026; Market Anarchy calculation

The $500 tier's exclusive burns 8x faster

Pro 500's one feature no other personal plan gets is Ultrafast, which OpenAI says generates tokens up to 8 times faster, around 300 tokens a second, and only on GPT-6 Astra for now. It draws your included allowance at 8 times the standard rate, then your credits at 6 times the standard credit rate (Nerdschalk, updated September 30, 2026).

Run the numbers. 25x Plus spent entirely on Ultrafast is 25 divided by 8: about 3.1 Plus allowances of Astra work. That's less than Pro 100's 5x at standard speed. Per Plus-sized unit of output, it's $160.

Speed has a price. For someone whose time costs more than $160 per Plus-sized block of output, Ultrafast can still pay. Faster tokens don't mean a task finishes eight times sooner, though, because a coding job also waits on tools, tests and you. Price it on the tasks you run.

The $2,500 in credits has a 63-day window

For an existing subscriber, the credit grant is the largest number in this change. $2,500 is 12.5 months of the $200 fee at face value. It expires on December 31, 2026 (MakeUseOf).

From October 30 to December 31 is 63 days. Those are the days the allowance is halved, so the credits work as a bridge: they cover the overflow above 10x for about two months. If you were using close to 20x, they buy time to measure where your usage sits. If you were using 5x, the cut doesn't reach you, and the credits are a bonus you'll likely leave unspent.

What to check before October 29

Three things decide what the cut costs you:

Which model your work runs on. Mostly Sol: your token budget barely moves. Mostly Astra: plan on half.
Your real multiple of Plus. Under 5x, Pro 100 costs half and covers it, at the same $20 per unit.
Whether you need Ultrafast. If you don't, Pro 500 is Pro 200 plus 15 more Plus allowances at list price.

We've seen this pattern before: in September, GitHub Copilot cut team AI credits by up to 44% and kept the price. Subscriptions that bundle compute get resized when the compute gets expensive. The price tag stays, and the number inside it moves.

What could prove this read wrong: OpenAI publishing a token table that weighs Astra differently from list price, or cutting Astra's API price. Either would close most of the gap between Sol and Astra users.