ChatGPT Pro 500 is the $500-a-month subscription OpenAI launched at DevDay on Tuesday 29 September 2026, and it arrived with a cut to the plan below it. The headline AIxploria’s AI Radar put on the story, “ChatGPT’s $500 Tier Buys Speed While the $200 Plan Shrinks”, is a fair summary. The new tier is the only personal plan with Astra Ultrafast, a mode that generates text up to eight times faster. The $200 plan keeps its price and loses half of its allowance.
We checked that report claim by claim against OpenAI’s own help pages, developer documentation, price tables and staff posts. Most of it holds. What the report does not show is the arithmetic underneath: after this change, every rung from Plus upward costs the same $20 for each unit of usage, so the extra money at the top pays for speed and nothing else.
This article sets out what ChatGPT Pro 500 includes, how the $200 plan changes, what Ultrafast does to an allowance and who should pay for it. We covered the leak in our ChatGPT Pro Max report and the earlier sign-up freeze in ChatGPT Pro subscriptions on hold. For the wider market, see our AI models and tools hub.
Table of contents
What ChatGPT Pro 500 Includes
OpenAI’s help article now opens with a plain statement: “ChatGPT Pro now offers Pro 500, a new $500/month plan that includes Astra Ultrafast.” Pro is no longer one plan. It is a family of three.
Three Pro plans, one exclusive feature
The help article on Pro tiers lists Pro 100 at $100, Pro 200 at $200 and Pro 500 at $500 a month. All three include Pro models, Codex, deep research, image creation, memory and file uploads. Only one line separates ChatGPT Pro 500 from the others: Ultrafast is “Included” on Pro 500 and “Not included” on the other two. Pro is billed monthly only, so there is no annual discount.
The ChatGPT Pro 500 allowance: 25 times Plus
OpenAI’s DevDay recap gives the size of ChatGPT Pro 500. “The new Pro 500 plan offers our highest usage allowance at 25 times the ChatGPT Plus allowance and includes access to Ultrafast,” it says. Thibault Sottiaux, whose X profile reads “Codex & ChatGPT @OpenAI”, set out the rest of the ladder the same day: Plus is 1x, Pro 100 is 5x and Pro 200 is 10x. The multiples describe usage in ChatGPT Work and Codex, which share one allowance.
| Plan | Monthly price | Allowance vs Plus | Ultrafast |
|---|---|---|---|
| Plus | $20 | 1x | No |
| Pro 100 | $100 | 5x | No |
| Pro 200 (old terms) | $200 | 20x | No |
| Pro 200 (new terms) | $200 | 10x | No |
| Pro 500 | $500 | 25x | Yes |
From Pro Max to ChatGPT Pro 500
ChatGPT Pro 500 leaked on 24 September under the name Pro Max. OpenAI’s own open-source Codex repository shows when the name changed. A commit dated 28 September, about 20 hours before the keynote, renamed three plan labels: “Pro Extra” became “Pro 200”, “Pro Standard” became “Pro 100” and “Pro Max” became “Pro 500”. The internal plan code is still ProMax. Naming the plans after their prices makes the ladder easier to read, and it also makes the cost of each step hard to miss.
How the $200 Plan Shrinks
The $200 plan is back on sale, and it is smaller. Both facts come from the same help article that introduces ChatGPT Pro 500.
From 20x to 10x
“New subscriptions that aren’t eligible for grandfathering include a lower usage allowance than previously offered with Pro 200 to reflect our increasingly efficient models,” OpenAI writes. “The monthly price remains $200.” Sottiaux gave the figure: the plan moves from 20 times the Plus allowance to 10 times. Engadget reported that in ChatGPT itself, the weekly cap on GPT-6 Pro messages also falls, from 200 to 100.
Who keeps the old allowance, and until when
Existing subscribers get a grace period. If a Pro 200 subscription was active at OpenAI’s eligibility cutoff, or in the seven days before it, the subscriber keeps the previous allowance “through Oct 29, 2026”. After that date the subscription moves to the lower allowance at the same price. OpenAI has not published the cutoff date, and says only affected users will receive an email. Keeping the old allowance does not add Ultrafast.
What Sottiaux told subscribers
Sottiaux announced the change on X at 06:41 UTC on 29 September, about ten hours before the keynote. “In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan,” he wrote. The post has been viewed more than 16 million times. He gave four reasons and promises:
- OpenAI is “committing to not reintroducing the 5h limit”, so weekly usage can be spent at any time.
- Subscribers will “always get more work done” over time as models become more efficient.
- OpenAI does not want “to artificially inflate the API list prices” to make subscriptions look generous.
- New features would arrive that “won’t draw on the usage”. That turned out to be Dots, the always-on agents.
In a second post he added that existing subscribers “will keep the 20X multiplier for a bit and also receive a lot of additional credits”.
The freeze that came first
The $200 plan had been closed to new buyers since 10 September. Sottiaux wrote that day that OpenAI was “going to pause subscriptions to our $200 Pro plan”, which “put the most strain on our systems” after the launch of GPT-6 Astra. The plan that reopened nineteen days later has the same name and price and half the allowance.
| Date (2026) | What happened |
|---|---|
| 13 August | OpenAI previews Ultrafast for GPT-5.6 Sol in the API |
| 10 September | New sign-ups for the $200 plan are paused |
| 24 September | A “Pro Max” plan leaks from ChatGPT’s code |
| 28 September | Codex code renames the plans Pro 100, Pro 200 and Pro 500 |
| 29 September | ChatGPT Pro 500 launches; the $200 plan reopens at 10x |
| 29 October | Last day of the old 20x allowance for eligible subscribers |
The ChatGPT Pro 500 Arithmetic: $20 per Unit on Every Rung
Put the prices and the multipliers side by side and a pattern appears that OpenAI has not spelled out.
The volume discount is gone
Divide each plan’s price by its allowance multiple. Plus is $20 for 1x. Pro 100 is $100 for 5x, or $20 per unit. The new Pro 200 is $200 for 10x, or $20 per unit. ChatGPT Pro 500 is $500 for 25x, again $20 per unit. Only the old $200 plan was different: $200 for 20x is $10 per unit. It was the one rung where buying more made each unit cheaper, and that discount is what OpenAI has removed.
What the extra money pays for
Twenty-five Plus subscriptions would cost 25 times $20, which is $500. On usage alone, ChatGPT Pro 500 is priced exactly like 25 Plus plans in one account. What it adds is the Pro feature set and the one thing no other personal plan can buy at any price: Ultrafast. OpenAI’s help article is explicit that “buying credits on Pro 100 or Pro 200 does not unlock Ultrafast” at launch. The headline is right in a literal sense. The $500 tier buys speed.
The yearly bill for ChatGPT Pro 500
Because Pro has no annual billing, ChatGPT Pro 500 costs $6,000 over twelve months. The $200 plan costs $2,400. The gap is $3,600 a year, paid for a larger allowance at the same unit price and for access to one speed mode.
What Ultrafast Does to a ChatGPT Pro 500 Allowance
Ultrafast is the reason to buy ChatGPT Pro 500, so its terms matter more than the multiple on the box.
Eight times faster, eight times the burn
OpenAI’s Speed documentation says GPT-6 Astra Ultrafast “generates tokens up to 8x faster than GPT-6 Astra in Standard mode in Codex”. The recap puts a number on it: 300 tokens per second. The same page sets the cost inside a subscription: “Ultrafast uses included subscription limits at 8x the Standard rate.” Fast mode, the cheaper option, uses limits at 2.5 times the Standard rate. OpenAI adds that these multipliers “don’t describe speed increases”, and that Ultrafast draws on included usage first and paid credits after that.
A 25x allowance behaves like 3x at full speed
Apply that burn rate to the ChatGPT Pro 500 allowance. If every task ran in Ultrafast, 25 units divided by 8 leaves about 3.1 units of Standard-equivalent work. In Fast mode, 25 divided by 2.5 leaves 10. A Pro 100 subscriber working at Standard speed has 5. This is an illustration, not a quota: OpenAI says actual usage depends on the model, context, reasoning and tools. But it shows the trade plainly. Ultrafast gets the same tokens out sooner. It does not buy more of them.
“Up to 8x” is token speed, not task speed
OpenAI is careful about what it measures. The comparison, it says, “measures token generation speed, not billing rates or overall task completion time”. A coding task also reads files, runs commands, waits for tests and waits for a person to review the result. None of that gets faster. The overall gain depends on how much of a task is spent generating. The table below applies an eight-fold speed-up to the generation share of a ten-minute task and leaves the rest unchanged.
| Share of the task spent generating | Ten-minute task becomes | Overall speed-up |
|---|---|---|
| 20% (2 minutes) | 8 minutes 15 seconds | 1.2x |
| 50% (5 minutes) | 5 minutes 38 seconds | 1.8x |
| 80% (8 minutes) | 3 minutes | 3.3x |
These are worked examples, not measurements. They show why ChatGPT Pro 500 suits people who spend their day waiting for long outputs and revising them, and why it does less for work dominated by builds and tests.
In the API, speed and price both rise six-fold
Developers do not need ChatGPT Pro 500 to reach Ultrafast. OpenAI’s API guide says the mode is “currently available to all API users at low rate limits”, and the recap describes it as “up to 6x” faster there. The price rises by the same factor. Rate limits start at 500,000 tokens a minute for usage tiers 1 to 3 and reach 5 million at tier 5.
| GPT-6 Astra, per million tokens | Standard | Ultrafast | Multiple |
|---|---|---|---|
| Input | $10 | $60 | 6x |
| Cached input | $1 | $6 | 6x |
| Output | $50 | $300 | 6x |
| Output, long context (over 272K input tokens) | $75 | $450 | 6x |
Where Ultrafast Came From
The speed mode behind ChatGPT Pro 500 did not start at DevDay. It has a seven-week history that explains some of its limits.
The August preview
On 13 August OpenAI published “Previewing Ultrafast mode: GPT-5.6 Sol at up to 14X the speed”. That preview ran on Cerebras hardware, generated “up to 750 output tokens per second” and was open only to “a select group of customers” through the API. OpenAI pitched it at incident response, financial research, customer support, commerce and live research. We reported the next step, a wider API rollout and a speed selector in the Playground, in OpenAI’s Ultrafast API expansion.
What changed for Astra
The DevDay version, the one inside ChatGPT Pro 500, runs a different model at different numbers. GPT-6 Astra Ultrafast reaches 300 tokens per second, against 750 for the GPT-5.6 Sol preview, and is “up to 8x” faster where the preview claimed 14 times. OpenAI’s recap and its Speed documentation do not name the hardware behind Astra Ultrafast. The recap does say that “GPT-6.1 Sol Ultrafast is coming soon”, which suggests the mode will spread to cheaper models.
Where Ultrafast is not available
Two limits apply even on ChatGPT Pro 500. In ChatGPT, Ultrafast “isn’t available to workspaces that require inference residency outside the United States”. In the API it “supports US data residency and global processing only” and does not support European or other regional endpoints. For Enterprise workspaces the mode is off by default until an owner enables it.
Fact-Checking the ChatGPT Pro 500 Headline
AIxploria’s AI Radar piece, by Sarah Nakamura, was published at 00:01 UTC on 30 September. We counted 18 checkable claims about ChatGPT Pro 500 in it and compared each with a primary source where one exists.
| AIxploria claim | What the sources say | Verdict |
|---|---|---|
| Pro 500 costs $500 a month | OpenAI help article | Confirmed |
| 25 times the Plus allowance | DevDay recap | Confirmed |
| Up to 8x faster in Codex, 300 tokens per second | Recap and Speed docs | Confirmed |
| Pro 200 allowance cut by 50%, from 20x to 10x | Sottiaux on X; Engadget | Confirmed |
| “Half the dollar in API spend” quote | Sottiaux’s post, word for word | Confirmed |
| API Ultrafast costs six times Standard: $60 in, $300 out | OpenAI API price table | Confirmed |
| Ultrafast drains the included quota faster | Pricing docs: 8x the Standard rate | Confirmed |
| August preview on Cerebras at 750 tokens per second | OpenAI post; it ran GPT-5.6 Sol, not Astra | Confirmed |
| 1.2 billion weekly ChatGPT users | DevDay recap | Confirmed |
| Weekly GPT-6 Pro messages fall from 200 to 100 | Engadget; not in the help article | Confirmed |
| One-time credit for current subscribers | Sottiaux; Engadget | Confirmed |
| Ultrafast also available through Enterprise or the API | Speed docs; API guide | Confirmed |
| No Ultrafast on Plus, Pro 100 or Pro 200 | Help article | Confirmed |
| The cut applies “from October 30” | True for grandfathered accounts; new buyers get 10x now | Partly |
| Every Pro subscriber gets a Dot outside the quota | Sottiaux confirms; Dots are limited to eligible markets | Partly |
| Pro 100’s usage against Plus is “n/a” | Sottiaux: Pro 100 is 5x | Incomplete |
| 35 million weekly builders in Work and Codex | Not in the recap or docs we read | Unverified |
| 100 GB of file storage | Not in the help article or pricing page | Unverified |
What holds up
Thirteen of the 18 claims are confirmed, twelve of them by OpenAI’s own pages or staff. The central claim, that ChatGPT Pro 500 sells speed while the $200 plan shrinks, is supported by three separate OpenAI sources: the help article, the recap and Sottiaux’s posts.
What is partial, missing or unverified
Two claims need a qualifier. The 30 October date applies only to subscribers OpenAI treats as grandfathered; anyone buying the $200 plan today gets the smaller allowance at once. And Dots are included with Pro only “in eligible markets”. Our report on Dots by OpenAI found that Pro subscribers in the UK, the EEA and Switzerland are excluded. One table entry is incomplete, and two figures, the 35 million builders and the 100 GB of storage, do not appear in the OpenAI pages we read.
Who Should Buy ChatGPT Pro 500
ChatGPT Pro 500 is narrow by design. It is worth the money for a specific kind of user and poor value for most others.
Developers who wait on generation
If you spend hours a day watching Codex write long diffs, reading them and asking for changes, waiting time is your main cost and Ultrafast removes most of it. Rationing matters, though. Use Ultrafast for the interactive loop, and Standard or Fast for long background jobs where nobody is waiting. Used that way, a ChatGPT Pro 500 allowance lasts far longer than the 3.1 units the all-Ultrafast case gives.
Heavy users who only want volume
If you want volume and not speed, the numbers are less kind. ChatGPT Pro 500 gives 2.5 times the allowance of the new $200 plan for 2.5 times the price. There is no saving per unit. Subscribers who relied on the old 20x allowance would need to spend $400 at the new rate to match it, and even the $500 plan gives them only a quarter more than they had for $200.
Teams and UK organisations
Companies should compare ChatGPT Pro 500 with the routes that do not need a personal subscription: the API, where Ultrafast is open to every account, and eligible Enterprise and Edu plans. UK and European organisations need to check residency first. Ultrafast is unavailable to workspaces that require inference outside the United States, which rules it out for firms with strict data-location or cybersecurity policies. US prices also exclude tax; at the UK’s 20% VAT rate, $500 becomes $600.
A checklist before you upgrade to ChatGPT Pro 500
- Measure how much of your working day is spent waiting for output. If it is small, Ultrafast will not change much.
- Check your current weekly usage in Settings before choosing a plan; the allowance multiples are relative to Plus.
- If you hold the old $200 allowance, note the date: it ends after 29 October 2026.
- Decide in advance which tasks justify Ultrafast, since it uses an allowance eight times faster.
- For teams, price the same workload on the API before buying several personal plans.
If you would like help comparing plans or building a usage policy, our AI strategy and IT consulting teams can work through the numbers with you.
ChatGPT Pro 500 FAQ
What is ChatGPT Pro 500?
It is OpenAI’s top personal subscription, launched on 29 September 2026 at $500 a month. It includes 25 times the ChatGPT Plus allowance and is the only personal plan with Astra Ultrafast.
How fast is Ultrafast on ChatGPT Pro 500?
OpenAI says GPT-6 Astra Ultrafast generates tokens up to eight times faster than Standard mode in Codex, at about 300 tokens per second. That measures generation speed, not the time a whole task takes.
Does Ultrafast use more of my allowance?
Yes. On ChatGPT Pro 500, Ultrafast uses included subscription limits at eight times the Standard rate. Once the allowance is used, it draws on purchased credits at six times the Standard rate.
What happened to the $200 plan?
It reopened to new subscribers on 29 September at the same price with half the allowance, 10 times Plus instead of 20. Eligible existing subscribers keep the old allowance through 29 October 2026.
Can I get Ultrafast without ChatGPT Pro 500?
Not on Plus, Pro 100 or Pro 200, even with extra credits. It is available through the OpenAI API at six times the Standard token price, and on eligible Enterprise and Edu plans.
Is ChatGPT Pro 500 better value than Pro 200?
Not per unit of usage. Both now work out at $20 for each multiple of the Plus allowance. The difference is the size of the allowance and access to Ultrafast.
References
ChatGPT’s $500 Tier Buys Speed While the $200 Plan Shrinks (AIxploria AI Radar)
About ChatGPT Pro tiers (OpenAI Help Center)
Speed: Fast and Ultrafast modes (OpenAI documentation)
Pricing, usage limits and credit rates (OpenAI documentation)
Ultrafast mode in the API (OpenAI developer docs)
API pricing (OpenAI developer docs)
Previewing Ultrafast mode: GPT-5.6 Sol at up to 14X the speed (OpenAI)
Thibault Sottiaux on the reopened $200 plan (X, 29 September 2026)
Thibault Sottiaux on the 1x, 5x and 10x ladder (X, 29 September 2026)
Thibault Sottiaux on pausing $200 sign-ups (X, 10 September 2026)
Update and centralize TUI subscription labels (openai/codex on GitHub)
OpenAI adds $500 Pro subscription, nerfs its existing $200 tier (Engadget)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
