The "Unlimited OpenAI API" Myth: What Subscriptions Really Give You
No unlimited OpenAI API exists at any price. What ChatGPT subscriptions actually provide: rolling usage windows, fair-use rules, and how to plan around them.
There is no unlimited OpenAI API. OpenAI does not sell one: the API is metered per token, and ChatGPT subscriptions enforce usage limits over rolling windows. Anyone promising unlimited OpenAI capacity, including vendors in our own category, is describing a limit you have not been shown yet. What subscriptions really give you is windowed capacity at a flat price, and that is valuable enough without the inflation.
This page exists because “unlimited” is the word that sells, and it deserves a debunking from inside the category that benefits from it.
Where the “unlimited” pitch comes from
Three sources keep the myth alive.
First, OpenAI’s own Pro marketing has used “unlimited” language for chat access to its flagship models, with a fair-use footnote attached. Second, gray-market resellers advertise “unlimited GPT API” backed by pooled accounts they control. Third, tools that turn ChatGPT subscriptions into endpoints, ours included, get compressed by third-party roundups into “unlimited API for $20” because that is the catchiest possible summary.
All three trace back to one true fact: a flat-price plan does not meter tokens. People hear “no meter” and repeat it as “no limit.” Those are different claims, and the difference is where every billing surprise in this category lives.
What a ChatGPT subscription actually gives you
A subscription buys a window that refills, not a meter that never stops.
ChatGPT plans include Codex, OpenAI’s coding agent, and Codex usage is governed by limits applied over rolling windows: hours-scale windows, with weekly components on some plans. The exact counts vary by plan and model and drift as models rotate, which is why we never reprint them; OpenAI publishes current figures on its pricing page. The full mechanics are in our Codex usage limits explainer.
What the windows are worth in API terms, using our planning estimates:
| Plan | Price | API-equivalent capacity (estimate) |
|---|---|---|
| ChatGPT Plus | $20/mo | ~$700/mo |
| ChatGPT Pro 5x | $100/mo | ~$3,500/mo |
| ChatGPT Pro 20x | $200/mo | ~$14,000/mo |
Treat the right column as a planning number, never a guarantee. The honest pitch for this category is “a $3,500 API workload for about $229 a month,” not “infinite tokens.” The wider survey of flat-price options is in our flat-rate OpenAI API roundup.
What OpenAI’s own “unlimited” means
OpenAI’s Pro positioning pairs the word with guardrails. The ChatGPT Pro help article prohibits “abusive usage, such as automatically or programmatically extracting data” and “reselling access or using ChatGPT to power third-party services.” Read together, “unlimited” means one person chatting as much as a person plausibly chats.
Codex is the carve-out that makes programmatic use legitimate at all: codex exec is the CLI’s documented non-interactive mode for scripts and CI (developers.openai.com/codex/noninteractive), and Codex is included in ChatGPT plans. That is intended functionality, but it runs inside plan limits, and OpenAI has the final call on how its services may be used. Documented programmatic access and unlimited programmatic access are not the same thing.
How to spot a fake “unlimited” offer
Four questions expose almost every inflated claim:
- Whose account does the work? If it is not your own subscription, capacity comes from accounts the seller pools, and pooled accounts are what OpenAI’s terms actually prohibit. Our look at API resellers walks that gray market.
- Is the mechanism named? “We have special access” is not a mechanism. “We run OpenAI’s Codex CLI signed into your account” is.
- Are the limits documented anywhere? A seller who never mentions limits has not removed them; they have hidden them.
- What happens at the limit? Throttling, silent queueing, and surprise cutoffs are all limits wearing costumes.
Any “unlimited” AI offer is a limit you have not been shown yet. The only variable is whether you find it in the docs or in production.
Our own product is not unlimited either
Codex Hosted runs OpenAI’s official Codex CLI signed in with your own ChatGPT account, so your workloads inherit your plan’s windows exactly as OpenAI enforces them. When a window exhausts, requests fail over to a second connected account if you have one, then to your own OpenAI API key, until the window resets. The request log shows which lane served every call, so the limit is visible, not mysterious. The full failover behavior is in what happens when you hit your Codex limit.
If you ever catch a page of ours saying “unlimited,” quote this article back at us. The product is windowed capacity with honest fallback, priced flat.
Plan with windows, not wishes
Sizing a subscription works like sizing bandwidth: measure your real monthly API spend, map it against the capacity estimates above, and leave headroom for bursts. Workloads under roughly $700 a month of API-equivalent usage fit a Plus window in our experience; steady multi-thousand-dollar workloads belong on Pro tiers or two accounts.
The calculator does this mapping against your actual bill in about thirty seconds. No infinity included, none needed.
Frequently asked questions
Is there an unlimited OpenAI API plan?
No. OpenAI's API bills per token with no flat tier, and ChatGPT subscriptions enforce usage limits over rolling windows. Any offer of unlimited OpenAI capacity hides a constraint somewhere: pooled accounts, throttling, or terms that let the seller cut you off.
Does ChatGPT Pro include unlimited API access?
ChatGPT Pro includes no API access at all; the OpenAI API is a separate metered product. Pro's marketing uses 'unlimited' for human chat usage subject to fair-use guardrails, and Codex usage under the plan has its own limits, published on OpenAI's pricing page.
What does 'unlimited' mean in ChatGPT Pro's marketing?
It refers to chat usage by one person, subject to OpenAI's abuse guardrails. The Pro terms prohibit programmatic data extraction, account sharing, and reselling access. It has never meant unmetered programmatic capacity.
Is ProxyLLM unlimited?
No. Codex Hosted runs on your ChatGPT plan's usage windows, which OpenAI sets and resets on a schedule. When a window is exhausted, requests fall back to a second connected account or your own API key until it resets. We publish the limits story instead of promising infinity.
How much API capacity does a ChatGPT subscription replace?
Our planning estimates: ChatGPT Plus at $20/month covers roughly $700 of API-equivalent work, Pro 5x at $100 roughly $3,500, and Pro 20x at $200 roughly $14,000. These are estimates from observed usage, not guarantees, and OpenAI adjusts limits over time.