ChatGPT Pro 500 costs $500 a month and adds Ultrafast Codex, up to 8x faster. Here is what it changes, what Pro 200 lost, and who should pay.
OpenAI added a $500 tier to ChatGPT on 29 September 2026. Pro 500 costs $500 a month, includes the highest usage allowance OpenAI sells, and is the only plan with Ultrafast, a faster service tier for GPT-6 Astra in ChatGPT Work and Codex. The same announcement reopened Pro $200 with a smaller allowance. If you searched for a "$500 Claude subscription", this is the plan you were probably looking for: as of 30 September 2026, Anthropic's top consumer tier is still Claude Max at $200.
What OpenAI launched
OpenAI announced Pro 500 at its DevDay event. It sits above Pro $100 and Pro $200 in the same Pro family.
| Plan | Price | Usage vs Plus | Ultrafast |
|---|---|---|---|
| Pro $100 | $100 a month | 5× | No |
| Pro $200 | $200 a month | 20× until 29 Oct, 10× from 30 Oct | No |
| Pro $500 | $500 a month | Highest of the three, multiple not published | Yes |
Three details from the launch are worth knowing before you compare it to anything else:
- OpenAI has not published a Pro 500 multiplier. Its wording is that Pro 500 "offers the highest included usage of the three plans". Any specific "25×" or similar figure is a third-party estimate, not an OpenAI number.
- Billing is monthly only. Annual plans are not offered on any ChatGPT tier, so the yearly cost is $6,000 only if you stay subscribed twelve months.
- It launched on the web. OpenAI said Pro 500 is sold through ChatGPT on the web at launch.
Otherwise the published Pro feature list is shared with Pro $200: deep research, agents, a 100 GB file library, memory, projects and scheduled tasks. Ultrafast is the headline difference.
Ultrafast: what "up to 8× faster" means
Ultrafast is a faster service tier for GPT-6 Astra. OpenAI says it delivers up to eight times the token-generation speed of Astra Standard in Codex, at up to about 300 tokens per second. Only Pro 500 can use it. Extra credits do not unlock it on Pro $100 or Pro $200, and it is unavailable to workspaces that need inference residency outside the US.
The 8× is a generation speed, not a finish-line speed. A Codex task is not all writing. It also runs tools, tests, builds and calls to outside services, and none of that gets faster. If a 10-minute task spends two minutes generating tokens, an 8× speedup cuts those two minutes to fifteen seconds. The task then takes about 8 minutes 15 seconds, not 1 minute 15 seconds. The more your work is waiting on the model to type, the more Ultrafast is worth.
It also drains the allowance faster. Ultrafast consumes your included plan usage at 8× the Standard rate. If you pay for it with purchased credits instead, the rate is 6×. So a Pro 500 subscriber who runs Ultrafast all day reaches the weekly limit sooner than one who stays on Standard. For context, Fast mode on GPT-6 Astra already costs 2.5× Standard in Codex and ChatGPT Work.
Codex and ChatGPT Work share one agent usage allowance, and OpenAI says there is no five-hour limit on Pro 200 or Pro 500. Weekly limits may still apply.
What happened to Pro $200
The same announcement changed the tier below it. OpenAI had paused new Pro $200 sign-ups and upgrades on 10 September 2026, citing demand for GPT-6 Astra. It reopened the plan on 29 September with a smaller allowance:
- Work and Codex usage falls from 20× to 10× Plus from 30 October 2026.
- GPT-6 Pro chat messages fall from 200 to 100 a week from the same date.
- Existing subscribers keep their current limits until 29 October, and reporting on the launch says they also receive a one-time grant of usage credits worth $2,500 that expires at the end of the year.
OpenAI attributed the reduction to its increasingly efficient models. The practical effect is a change in what $200 buys. After 30 October, Pro $100 (5×) and Pro $200 (10×) cost the same per multiple of Plus, $20 each. For a fuller three-way comparison, see ChatGPT Pro $100 vs $200 vs $500: Usage Limits Explained.
Who should pay $500
Pay for Pro 500 if you sit in Codex or ChatGPT Work for hours, you watch the output stream, and the wait for tokens is the slowest part of your day. The clearest case is interactive coding where the model writes a lot and tools run quickly. It also makes sense when the time is billed to a client, since $500 is a fraction of one saved workday.
Stay on Pro $200 or $100 if your tasks are dominated by test runs, builds and long-running agents, or if the weekly allowance, not speed, is what stops you. Ultrafast's 8× burn rate makes that problem worse, and OpenAI has not said what the extra allowance on Pro 500 is worth in multiples, so you cannot price it per unit the way you can for the cheaper tiers.
A simple test before committing: log a week of Codex sessions and note how much wall-clock time is the model generating versus everything else. If generation is under a third of the total, Ultrafast will not change your day. Because the plan is monthly, you can try it for one cycle and drop it at renewal.
What about Claude?
There is no $500 Claude plan. Anthropic sells Claude Max at two prices, $100 for 5× and $200 for 20× the Pro allowance, and as of 30 September 2026 nothing sits above that in the consumer lineup. If you want more Claude than Max 20× gives you, the routes are extra usage on top of the subscription or the API, not a bigger plan. Our guides cover each side:
- Claude Max 100 vs 200: Same Features, 5× vs 20× Usage
- Claude Code Pricing & Token Economics: Subscription vs. Pay-As-You-Go API
- Cursor Ultra vs. Claude Max 20x vs. ChatGPT Pro: The $200/Month Developer Tier Showdown
The new $500 price also sets OpenAI's ceiling at 2.5 times the $200 that Anthropic, Cursor and others treat as the top individual tier. Whether Anthropic follows is speculation. Nothing has been announced.
Bottom line
Pro 500 is a speed product with a bigger allowance attached. Ultrafast delivers up to 8× faster Codex token generation, uses allowance at 8× the Standard rate, and is exclusive to the $500 tier. It is worth the money for people who spend their day waiting on the model to write, and poor value for anyone whose limit is the weekly allowance or whose work is mostly tool execution. Pro $200 stays open, but from 30 October it holds half its old allowance. Claude has no matching $500 plan. Check OpenAI's Help Center tier page before paying, because limits and plan details have changed several times in the past three weeks.