ChatGPT Pro Max surfaces at $500 with "Fastest Work and Codex"
TestingCatalog broke the story. It reported that the plan has surfaced in unreleased ChatGPT subscription references, where its description differs from the existing Pro tier mainly through access to the "Fastest Work and Codex." Researcher Tibor Blaho posted the same find on X that day. BleepingComputer followed up, reporting that the unannounced plan has started appearing in ChatGPT's subscription interface alongside Plus and existing Pro options.
The price seen so far varies by tax. According to BleepingComputer, one version of the page lists Pro Max at $500 per month, while another shows $600 after local taxes or VAT are included. TestingCatalog put the $600 figure down to VAT. So $500 is the base US-style figure, and $600 is a tax-inclusive display, not a separate higher price.
BleepingComputer also says the listing mentions access to the company's most capable frontier models, maximum memory, 100GB of file storage, and early access to new features. Those details come from one outlet reading a pre-release screen. Treat them as draft marketing copy, not confirmed entitlements. OpenAI's public pricing page still doesn't list Pro Max; it lists only six plans: Free, Go, Plus, Pro, Business, and Enterprise.
OpenAI's own Codex repository adds a "promax" plan
The best evidence so far comes from OpenAI's own code, not a screenshot. According to Progressive Robot, at 00:10 UTC on 25 September, a commit by Eric Traut was added to the public openai/codex repository titled "Add Pro Max plan support and update Pro display names". It was merged as pull request 47971. An Italian tech blog by Pasquale Pillitteri separately reported that a promax plan, shown as "Pro (Max)", has existed in Codex's public repository since September 25, 2026 and has a price in each of the 14 checkout configurations checked, from $500 in the United States to €515 in Italy.
A commit that teaches the Codex client to recognise a plan tier is a strong sign the tier is real and close. It still isn't an announcement, and features, name and price could all change before launch.
The commit also hints at how Codex will treat Pro Max accounts. Progressive Robot reports that the plan qualifies for "experimental context", a feature the Codex code limits to Plus, Pro, Pro Lite and now Pro Max accounts signed in with ChatGPT. More importantly, when a Pro Max user hits a limit, Codex shows the same message it shows Pro users: "You've hit your usage limit. Visit to purchase more credits". The outlet's conclusion is that even at $500 a month, the code assumes ChatGPT Pro Max users can run out and will be pointed to buy credits. Nothing in the commit suggests unlimited use.
For anyone hoping $500 buys unlimited Codex, that's the most practical detail here. Based on the client code, Pro Max will still have a usage ceiling, just like the current Pro tiers.
What the $100 and $200 Pro tiers already sell
Today's lineup explains why a speed-first plan stands out. OpenAI's help center says ChatGPT Pro comes in $100 and $200 tiers with the same core features. The main difference is usage: Pro $100 gives 5x the usage of Plus, and Pro $200 gives 20x. Every Pro tier includes Pro models, Codex, deep research, image creation, memory and file uploads. OpenAI also says some models have their own usage allowances, which can differ between the two tiers.
Codex draws on the same pool. OpenAI's Codex pricing page says Codex is included in your ChatGPT Free, Go, Plus, Pro, Business, or Enterprise plan and that local messages and cloud chats share your plan's usage allowance. Weekly limits may also apply. Current limits appear in the Codex usage dashboard. OpenAI's help center says Support won't reset ChatGPT or Codex limits. Once you hit a limit, you wait for the reset or use another option shown in your account.
Until now, moving up a tier bought more quota at the same speed. OpenAI's public pricing grid lists response times as "Fast" for both Plus and Pro. Pro Max's one-line pitch, "Fastest," would be the first time the company sold speed as a separate tier feature. Pillitteri's blog argues that Pro Max would be the first consumer subscription to sell response speed as a line item of its own, at least on paper. That fits the evidence, but it's still analysis. OpenAI hasn't said whether Pro Max also raises quotas. Higher usage limits may also be part of the package, although that is not confirmed yet, TestingCatalog notes.
The Pro $200 pause shapes who can buy what
Pro Max is showing up while OpenAI is limiting its current top plan. The help center says that since September 10, 2026, OpenAI has temporarily paused new sign-ups and upgrades to Pro $200 (Pro 20x). That includes upgrades from Free, Go, Plus and Pro $100. Existing Pro $200 subscriptions keep renewing, and Pro $100 is still open to new and existing subscribers. OpenAI calls Pro $200 its highest usage tier during the pause.
The rules for current $200 subscribers are strict, and anyone thinking about switching plans should know them first:
- If you cancel or schedule a downgrade from Pro $200, you keep it until your billing cycle ends and can undo the change in billing settings before then.
- If your Pro $200 access ends, you may get a one-time return within 30 days. That applies if you had Pro $200 on September 10, 2026, or your access ended in the 30 days before that date, and you haven't already used the return.
- To use the return, sign in to the same account and go to the ChatGPT pricing page. If you're eligible, Pro $200 will be available to buy there.
- A refund that cancels Pro $200 ends access immediately and moves the account to Free. New Pro $200 promo redemptions are paused.
TestingCatalog says the pause is because of system load. That context matters for reading Pro Max. The inference is that if OpenAI can't take on more heavy users at $200, a $500 tier helps it charge more for the most compute-heavy customers. OpenAI hasn't said Pro Max replaces Pro $200 or opens a way around the pause. Nothing in the evidence suggests current $200 subscribers will be moved to it.
Cerebras is the likely engine behind "Fastest," but that's unconfirmed
Both outlets that first covered the plan think the speed may come from Cerebras hardware. TestingCatalog notes that OpenAI already uses Cerebras for its latency-focused inference path, including GPT-5.6 Sol Ultrafast. This makes it plausible that the "Fastest" tier could rely on additional Cerebras capacity, though OpenAI has not confirmed a direct connection between Cerebras and Pro Max.
The partnership is real. OpenAI announced it on January 14, 2026, saying it would add 750MW of "ultra low-latency AI compute" to its platform. OpenAI described Cerebras systems as built to speed up long model outputs by putting compute, memory and bandwidth on one giant chip. It said the capacity would be added to its inference stack in phases across workloads and come online in tranches through 2028.
BleepingComputer points to earlier deployments: Codex-Spark launched on Cerebras at more than 1,000 tokens per second, and OpenAI later demonstrated GPT-5.6 Sol running at up to 750 output tokens per second in its Cerebras-powered Ultrafast mode. Those are OpenAI's own published figures for specific models and modes. They don't tell us how fast Pro Max will be, and BleepingComputer says neither OpenAI nor Cerebras has confirmed that connection.
The workload makes the speed pitch believable. TestingCatalog notes that Work is designed for longer tasks, while Codex can operate on coding jobs that continue for hours or even days, making inference speed much more valuable in these scenarios. For an agent working through a large refactor, faster token generation shortens the time until results are ready to review. It doesn't make the model smarter. Speeding up long multi-step agent runs is the one area where a speed-priced tier could clearly pay off in saved hours.
Pro Max against Claude Max, Cursor and Google AI Ultra
At the leaked price, Pro Max would be the most expensive mainstream single-user AI subscription. TestingCatalog says that if released in its current form, Pro Max would sit far above most individual AI subscriptions, including $200 plans from Anthropic and Cursor and Google's roughly $250 AI Ultra tier. Over a year, that's $6,000 per seat before tax. Monthly billing is the only option: OpenAI's help center says it doesn't offer annual billing for Go, Plus or Pro.
The price comparison doesn't tell the whole story. As Pillitteri points out, Anthropic's $200 plan mostly buys usage capacity, not a fast lane, and the same goes for much of the competition. A team comparing these plans is comparing different things: quota on one side, and on OpenAI's side, possibly speed plus a quota nobody has disclosed. Until OpenAI publishes multipliers and some measure of how much faster "Fastest" is, you can't work out cost per task.
OpenAI has an obvious venue to make it official. TestingCatalog notes that OpenAI DevDay takes place on September 29, where the company is expected to cover developer tooling and new subscription tiers. That's a reasonable guess about timing. OpenAI hasn't committed to announcing Pro Max there.
What this means for you
For now, there's nothing to buy. Pro Max isn't on sale, so the decision is whether to wait for its terms or keep your current plan. It's worth watching mainly if you're a heavy individual Codex or ChatGPT Work user who regularly hits Pro limits or waits on long agent runs. Most Plus users and light Pro users have no reason to care yet. Business and Enterprise admins should note that this is a personal-plan tier. OpenAI's help center says personal Pro subscriptions and Business workspace subscriptions are separate, and nothing reported so far puts Pro Max into workspace plans.
- If you're on Pro $200, don't cancel or downgrade expecting to move to Pro Max. New $200 sign-ups are paused, and you get only one return within 30 days of losing access.
- If you need more than Plus right now, Pro $100 is open. It gives 5x Plus usage, and an upgrade takes effect immediately.
- Before paying $500, wait for OpenAI to publish Pro Max's usage allowance and a concrete speed claim. According to the Codex client code, Pro Max accounts still hit limits and get pointed to buy credits.
- Check the Codex usage dashboard to see how often you hit weekly or model-specific limits. That tells you whether you need more quota or more speed.
- Budget Pro Max at $6,000 a year per seat before tax. Displayed prices so far range from $500 to $600 depending on VAT.
OpenAI has now put Pro Max in both its web app and its open-source Codex client, so a launch looks likely, possibly at DevDay on September 29. When it's announced, check two numbers: the usage multiplier over Plus and the speed advantage for Codex and Work sessions. With both in hand, developers can compare Pro Max's cost per finished task directly against Pro $200, Claude Max and Google AI Ultra.