ChatGPT Pro 500: What $500 a Month Buys, and What Changed on 29 October

OpenAI launched a $500-a-month ChatGPT plan on 29 September 2026. It's called Pro 500, it sits above the existing $100 and $200 Pro tiers, and it adds two things: the largest usage allowance OpenAI sells to one person, and access to Ultrafast, a faster version of its newest model.

That was the headline at launch, and dividing the price by the usage gave the more useful finding: Pro 500 charged exactly what Plus charges per unit of usage, and the one discounted rung, Pro 200, had its allowance cut for new subscribers the same day. A month later OpenAI changed both halves of that. On 29 October it doubled the Work and Codex allowance on Pro 100 and Pro 500, left Plus alone, and cancelled the Pro 200 cut. The day before, it added API credits worth the plan price to all three Pro tiers.

Everything below comes from OpenAI's own pages: the DevDay recap, the help article on Pro tiers, the ChatGPT release notes, the pricing page, the help article on usage in ChatGPT Work and Codex, and the one on API credits. I read them on launch day and again on 3 November 2026.

The Short Answer

Plan	Price	Usage, per OpenAI	Ultrafast	Monthly API credits
Plus	$20/mo	The baseline, unchanged on 29 October	No	None
Pro 100	$100/mo	Work and Codex allowance doubled on 29 October	No	$100
Pro 200	$200/mo	"More usage than Pro 100". The September cut is cancelled	No	$200
Pro 500	$500/mo	"The highest included usage of the three plans". Work and Codex doubled on 29 October	Yes	$500

OpenAI no longer prints a multiple for any Pro tier. Its help centre says only that "Pro 200 includes more usage than Pro 100" and that "Pro 500 offers the highest included usage of the three plans". At launch the DevDay recap put Pro 500 at "25 times the ChatGPT Plus allowance", and OpenAI's usage tables labelled the older tiers "Pro 5x" and "Pro 20x". I found none of those labels in the help centre on 3 November, and the 29 October doubling means the launch figures no longer describe Work and Codex anyway.

So Pro 500 is for exactly one kind of person: someone whose work already stalls at Pro 200's limits, or who loses real money waiting for long agent runs to finish. Since October there is a second kind: a developer who already spends $500 a month on OpenAI's API. Everyone else is better off on a cheaper rung.

What $500 Buys

The Largest Allowance OpenAI Sells

This is the main thing you're paying for. The DevDay recap put a number on it at launch, 25 times Plus. On 29 October OpenAI doubled the Work and Codex part of it, for Pro 100 as well, and said Plus usage is unchanged. It didn't publish a new multiple, and it warned subscribers that "a larger allowance does not guarantee twice as many completed tasks."

Two limits on what any of this means. It isn't a message count. OpenAI's usage guide for ChatGPT Work and Codex says your allowance "does not correspond to a fixed number of messages or tasks", and what it covers depends on the model, the task and the settings. And it isn't unlimited. Every model on Pro keeps its own allowance, and at a model's limit, OpenAI says, "it may be unavailable". Everyday text chat is listed as unlimited on every ChatGPT plan, Free included, so that part of the product doesn't change as you climb the ladder. OpenAI also says Pro 100 and Pro 500 keep their current weekly GPT-6 Pro message limits in chat. The doubling was for Work and Codex.

One more change is coming. OpenAI's usage article says that from December, the Plus, Pro and Business allowances will be shared across ChatGPT and Codex and measured in tokens, with instant chat not drawing from it. Expect the way these tiers are described to shift again then.

Ultrafast

Ultrafast is OpenAI's "premium speed tier". The recap describes it as "up to 8× faster token generation (300 tokens per second) in Codex". On ChatGPT plans it runs OpenAI's newest models, GPT-6 Astra and GPT-6.1 Sol, inside ChatGPT Work and Codex, and you choose it from the model picker.

Three details matter before you pay for it:

• It's only on Pro 500 among personal plans. OpenAI's help article says that "buying credits on Pro 100 or Pro 200 does not unlock Ultrafast".
• It spends your allowance fast, then your credits. Ultrafast uses "your plan's included usage first at 8× the Standard rate for the selected model", then draws from your credit balance at six times the standard credit rate. Faster work runs through the allowance faster too.
• It lives in Work and Codex, not in ordinary chat. If most of what you do in ChatGPT is conversation, Ultrafast doesn't touch it.

API Credits Worth the Plan Price

Since 28 October all three Pro tiers include credits for OpenAI's API: $100 a month on Pro 100, $200 on Pro 200 and $500 on Pro 500. You claim them after seven days on the plan into one linked API organisation, and they don't add to your ChatGPT or Codex allowance. They're for building your own apps and agents. Anthropic did the same for Claude Max three weeks earlier, and I compared the two ladders in Claude Pro vs Max vs Ultra.

Everything Else Is Shared

The rest of the Pro card applies to all three tiers: OpenAI's Pro reasoning model, maximum deep research, the largest memory and context, and dots, OpenAI's always-on agents. OpenAI's pricing grid lists one set of context figures for Pro as a whole, so nothing there suggests Pro 500 reads more of a document at once than Pro 100. It gets more usage, more speed and more API credit.

The Per-Unit Math, Before and After 29 October

At launch, the ladder divided out like this, using OpenAI's own multiples:

• Plus: $20 for 1x. $20 per unit.
• Pro 100: $100 for 5x. $20 per unit.
• Pro 200: $200 for 20x. $10 per unit.
• Pro 500: $500 for 25x. $20 per unit.

Three of the four rungs charged the same rate. Only Pro 200 was cheaper per unit, at half price, and OpenAI cut its allowance for new subscribers that same day.

The 29 October change breaks that table for Work and Codex. Pro 100 and Pro 500 now get twice their previous Work and Codex allowance, Pro 200 subscribers on the lower allowance were brought up to the higher one, and Plus didn't move. OpenAI hasn't published new multiples, so I won't print a new rate as if it had. But if the launch figures were right, doubling them puts Pro 100 and Pro 500 at about half of Plus's rate for Work and Codex, the same rate Pro 200 always had. For agent work, every Pro rung is now a bulk discount on Plus. For chat, OpenAI says the weekly GPT-6 Pro message limits on Pro 100 and Pro 500 didn't change.

Anthropic went the other way. When it renamed Max 20x to Ultra at the end of October, its $200 rung went from twenty times Pro's session usage to ten times Pro's weekly usage, so every Claude rung now charges Pro's rate. The two ladders still sit at the same $100 and $200 prices, but the per-unit picture has flipped.

Now look at the step from Pro 200 to Pro 500 on its own. You pay $300 more. That buys the largest allowance OpenAI sells, Ultrafast, and $300 more a month in API credits. For a developer who would have spent that $300 on OpenAI's API anyway, the step costs nothing extra. For anyone who never opens the API dashboard, the credits are worth nothing, and the price is for headroom and speed.

What Changed for Pro 200

The $200 tier had a strange autumn. OpenAI paused new sign-ups on 10 September, reopened them on 29 September with a smaller allowance for new subscribers, and let existing subscribers keep the old allowance until 29 October. On 29 October, the day that grace period was due to end, it cancelled the cut. In OpenAI's words: "The reduction announced in our September 29 email is canceled. If you already had the higher Work and Codex allowance, you keep it. If you had the lower allowance, it has doubled to match."

The details:

• New Pro 200 subscriptions get the higher allowance, and the price stays $200.
• On Pro 200, the lower weekly GPT-6 Pro message limit now matches the higher one.
• If you cancel and subscribe again later, you get the same allowance as every other Pro 200 subscriber.
• None of this adds Ultrafast, which stays on Pro 500.

So if you held off on Pro 200 in October because of the smaller allowance, that reason is gone.

"Unlimited", According to the Pro Card

OpenAI's Pro card says "Unlimited subject to abuse guardrails" under its feature list. The help centre is plainer: every model on Pro has its own allowance, and at a model's limit "it may be unavailable" until that allowance resets.

That gap matters most at $500. If you're paying for headroom, find the limit you actually hit before you pay, because it's a per-model limit. Buying a bigger allowance only helps if that's the specific wall you keep hitting.

Who Should Pay $500

• Pay it if you run long jobs in ChatGPT Work or Codex every day, you already hit Pro 200's limits, and the waiting costs you more than $300 a month. Ultrafast is the only thing on the ladder that buys speed rather than volume.
• Pay it if you already spend about $500 a month on OpenAI's API. Pro 500 now returns its price as API credits, so the ChatGPT plan comes almost free.
• Stay on Pro 200 if you rarely hit its limits. The September cut is cancelled, so there's no deadline to beat.
• Stay on Pro 100 or Plus if you mostly chat. Everyday chat is unlimited on every tier, and the $500 plan's extras sit in Work and Codex.
• Consider a team seat instead only if other people need to share your work. For one person, a Premium Business seat still needs a second paid seat, and it changes who owns your chat history.

The full ladder, from Free and the $6 Go plan up, is in ChatGPT Go vs Plus vs Pro, and every tier's price, including Business, Enterprise and the API, is in how much ChatGPT costs.

Before You Buy More of One Vendor

Most people who look at a $500 plan got there by running into limits on a $20 one. Before paying for the largest allowance OpenAI sells, it's worth asking which work actually needs OpenAI's top model, and which of it just needs a good model. ChatGPT's limits sit on specific features such as uploads, images, research and the Pro models, and a lot of everyday drafting and checking runs just as well on another lab's model.

That's what izzedo chat is for. It puts OpenAI's GPT models in one thread with Claude, Gemini, Grok, DeepSeek and more, so the jobs that don't need OpenAI's top tier go somewhere else, and a second model can check the ones that do. Paid plans start at $6 a month, and the free plan needs no card. It won't give you Ultrafast. For most people, though, the problem isn't speed. It's that one vendor's limits are standing between them and the work.

The Bottom Line

ChatGPT Pro 500 costs $500 a month for the largest allowance OpenAI sells, Ultrafast, and since 28 October $500 a month in API credits. At launch it charged exactly Plus's rate per unit. Since OpenAI doubled the Work and Codex allowance on 29 October it's cheaper than that for agent work, though OpenAI hasn't said by how much. It's worth it if speed on long agent jobs is what you're short of, or if you already spend that much on OpenAI's API. If you're short of room on everyday work, you're paying the top rung's premium to solve a problem that a cheaper plan, or a second model, would fix.

---

Hitting ChatGPT's limits on work that doesn't need OpenAI's top model? Try izzedo chat free. No credit card required.