ChatGPT API credits are the monthly OpenAI API credits included with qualifying ChatGPT Pro and Business subscriptions, granted to one API organization owned by the same person and spent through that organization’s project API keys. They pay for OpenAI API usage until the allowance runs out, and whatever remains expires when the billing cycle ends.
OpenAI’s changelog entry for October 28, 2026 reads: “Eligible ChatGPT Pro and Business subscribers can now claim up to $500 per month in free API credits as a subscription benefit.” A company already paying for ChatGPT Business, and building apps, agents or an AI workspace on its own OpenAI key, will want to know whether the plan can fund that key. It can, within four rules: who may claim, which organization receives the grant, what the benefit refuses to cover, and what every request returns once the money is gone. The table below is the short version, and each row gets a section after it. Anthropic’s equivalent has different rules; see Claude API credits.
ChatGPT API credits at a glance
Up to $500 a month per subscription, sent to one API organization, spent with that organization’s keys, and gone at the next renewal. Every row comes from OpenAI’s guide to API credits for ChatGPT subscribers.
| Question | OpenAI’s rule | What to plan for |
|---|---|---|
| How much? | Pro: $100, $200 or $500 a month, by tier. Business: $20 or $100 per paid seat, by seat type | A Business workspace gets one grant, “capped at $500 per workspace per month” |
| Who can claim? | Someone who owns both the ChatGPT subscription or Business workspace and the receiving API organization, with one email address on both accounts | The paid subscription must be at least seven days old |
| Where does it go? | The one API organization chosen at the claim | Changing it means contacting OpenAI support |
| How is it spent? | Requests made with one of that organization’s project API keys | Keys from your other organizations never touch it |
| What does it cover? | Paid API Platform usage billed to that organization | Not Codex through API keys, and no lift to ChatGPT or signed-in Codex limits |
| When does it expire? | ”Unused credits expire at the end of the billing cycle and don’t roll over” | Annual upfront Business plans receive the whole year in one issuance |
| What happens at zero? | Anything you bought takes over; with nothing left, requests fail with HTTP 429 | The error code says the money ran out, not that a rate limit was hit |
How much does each ChatGPT plan include?
Pro gives a fixed amount by tier; Business earns per paid seat into one workspace total that stops at $500 a month. Amounts are in US dollars:
| ChatGPT subscription | Each month | How it’s counted |
|---|---|---|
| Pro ($100 tier) | $100 | One subscription |
| Pro ($200 tier) | $200 | One subscription |
| Pro ($500 tier) | $500 | One subscription |
| Business Standard | $20 per paid Standard seat | Added up across the workspace, to at most $500 |
| Business Premium | $100 per paid Premium seat | Same total, same ceiling |
Four notes on that table:
- The ceiling arrives early. $500 is 25 Standard or five Premium seats; anything past that adds nothing, and “reassigning an existing seat to another person doesn’t create extra credits.”
- The first grant is partial. It is “prorated for the time left in your current subscription billing cycle” and expires with it. A Pro upgrade, or a Business workspace that adds or upgrades seats, can top it up partway through on the same prorated basis; reductions count from the next renewal.
- Annual upfront Business plans get the year at once. They “receive the full year’s grant in one issuance, and the grant expires at the end of the annual billing cycle,” with a shared workspace ceiling of $6,000 for that year.
- Some plans never qualify. Free trials, complimentary subscriptions and ChatGPT Enterprise are out. Dropping to Plus, Go or Free, or moving from Business to Enterprise, removes the organization link at the end of the billing period. Cancelling stops new grants without revoking what was already issued, which stays usable until it expires.
Who can claim, and which organization receives the grant?
One person must own both sides: the ChatGPT subscription or Business workspace, and the API organization that receives the grant, with the same email address on each account. In a company, that means the Business workspace owner who claims must also be an owner of the API organization whose keys your software uses. If those are two different people today, one needs owner access on the other side before the claim goes through.
Either product can start it, and “both paths link the same subscription benefit to your selected API organization.”
| Start in | Where to go | What you confirm |
|---|---|---|
| ChatGPT | Pro: Settings → Billing. Business: the workspace’s billing settings in ChatGPT or the Admin Console | The benefit, and which of your API organizations receives it |
| API Platform | Sign in with the subscription’s email address, pick the API organization you own, open Billing | The Pro subscription or Business workspace you own |
Afterwards the grant shows in API Platform billing settings, and no API payment method is needed to start: “You can use the API with active free credits.”
Three things to settle before confirming:
- Pick the organization you’ll keep. OpenAI’s instruction is to “contact support if you need to change the linked API Organization.” A newcomer to the API can link a Personal API Organization, which suits one person’s Pro experiments but not a company whose apps share an organization.
- Call the API with that organization’s keys. The guide says “grants apply only to the linked API organization” and to “select that organization and use one of its project API keys when making requests.” It’s silent on how projects inside the organization share the grant, so when one workload needs its own ceiling, OpenAI’s documented control is a project hard spend limit, covered in AI API spend limits when you bring your own key.
- Switching plans breaks the link. Moving from Pro to Business ends the Pro link; a Business workspace owner can create a new one after the workspace’s paid subscription has been active for seven days.
What do ChatGPT API credits cover?
“Paid API Platform usage billed to the linked organization,” with Codex as the one exclusion OpenAI spells out. The guide’s list:
- the Responses API, for text, code and image understanding
- the Agents API, including OpenAI-hosted sandbox charges
- the Decisions API, for structured decisions
- GPT-Live and the Realtime API, for real-time voice
- the Images API, for image generation and editing
- built-in tools such as web search, file search and Code Interpreter, with their usage and storage charges
The exclusion reads: “Codex usage through API keys is not supported by this benefit. These API credits also don’t increase your ChatGPT or signed-in Codex usage limits.” A developer signed in to Codex through ChatGPT keeps exactly the plan’s limits; the grant can’t raise them. The benefit also stays “subject to API terms and policies including API rate limits,” so a busy app can be throttled with money still unspent.
What happens when ChatGPT API credits run out?
The free grant is spent before anything you bought; after it, bought credit carries on, and an organization with neither gets HTTP 429 with the error code credit_balance_exhausted.
| What the organization holds | What OpenAI does | What your code sees |
|---|---|---|
| Bought credit remains | ”API usage continues using your purchased credits” | No error, but every request now costs money you paid |
| Nothing remains | ”API requests fail with HTTP 429 and error code credit_balance_exhausted” | Every request fails until someone tops up or the subscription renews |
That 429 is easy to misread, because OpenAI returns the same status when a spend cap is reached. Its spend-limits guide says to “check error.code” to tell the causes apart:
error.code | What it means | What restores traffic |
|---|---|---|
credit_balance_exhausted | The organization has nothing left to spend | ”Add credits,” or wait for the next grant |
organization_spend_limit_exceeded | An organization hard limit you set was reached | Raise or remove the limit, or wait for the monthly reset |
project_spend_limit_exceeded | A project hard limit you set was reached | The same, for that project |
organization_usage_limit_exceeded | OpenAI’s approved usage limit was reached | Request a higher approved usage limit |
Retry logic that backs off on every 429 as if it were a rate limit will keep retrying an account that won’t refill before renewal. OpenAI’s remedies are to buy more in the organization’s billing settings or to “set up Auto Recharge to add paid credits to continue building without errors and interruptions.” Either turns a free allowance into a paid one once it’s spent. Decide in advance, per workload, whether to keep running at your own expense, fail over to another provider, or stop until renewal, and make sure someone watches what’s left in API Platform billing settings.
How Insulin uses the grant
Insulin’s organization-level OpenAI connection takes an ordinary API key, so a project key from the receiving organization lets the grant pay for what organization agents send to OpenAI, with billing “handled directly by OpenAI.” An org ADMIN connects it under Settings → Integrations; the OpenAI integration also takes an optional Organization ID “if you belong to multiple organizations,” which should then be the receiving one. A non-empty AI Gateway allow-list must include OpenAI, and an organization agent reaches OpenAI first when an OpenAI model is its Default model. Changing the key means deleting the integration and creating a new one.
The built-in assistant and personal agents never reach that key. Per Model Selection, they add the providers each member connected and “never reach the organization’s keys,” and the built-in assistant is also what members talk to in Slack and Microsoft Teams. A member’s own OpenAI connection is a Codex sign-in, which the grant doesn’t cover. So the grant pays for organization agents, whoever asks them and whether they run in a chat, a channel or a job, while the assistant, Slack, Teams and personal agents run on members’ own connections or the Fours-hosted models.
At zero, a failed call “retries on your next connected provider,” and Insulin leaves the rest to you. The Fours-hosted models are “the fallback the turn degrades to” unless the AI Gateway rules them out (what the AI Gateway controls), in which case the turn tries another organization key or stops and says why; a turn that switched models and succeeded ends with an italic note naming both (When a model fails over). An API-key connection never shows the Connection expired alert (When a connection expires), so OpenAI’s billing settings, not the integration card, show the grant running low. And the grant pays OpenAI only: on your own key Fours charges “a flat fee per unit” from the Insulin balance, and “a zero balance still pauses BYOK requests,” so on pay as you go a workspace can stop with part of the ChatGPT grant unspent; monthly billing is never stopped for a balance. Insulin credit keeps its own clock: monthly free credit expires at the end of the month it was issued in, and paid credit lasts three years.
Whether to run on your own OpenAI key at all, grant or no grant, is weighed in BYOK vs managed models for enterprise AI.
Sources
Primary sources for the platform rules cited above. Last verified November 3, 2026. Providers change their products, prices and terms without notice — check the source before relying on a detail.
- API credits for ChatGPT subscribers | OpenAI API — Pro $100, $200 and $500 tiers at $100, $200 and $500 a month; Business $20 per paid Standard seat and $100 per paid Premium seat, pooled and capped at $500 per workspace per month; first grant prorated; annual upfront Business plans get the full year's grant in one issuance, expiring at the end of the annual cycle, with a $6,000 annual workspace cap; eligibility: an active, paid Pro or Business subscription at least seven days old, no free trials or complimentary subscriptions, an owner of both the subscription or workspace and the receiving API organization, the same email address on both accounts; both claim paths, from ChatGPT and from API Platform; no API payment method required; covers paid API Platform usage billed to the linked organization, Codex through API keys not supported and no increase to ChatGPT or signed-in Codex limits; subject to API rate limits; unused credits expire at the end of the billing cycle and don't roll over; free credits spent before purchased; with no credits left, HTTP 429 and credit_balance_exhausted; grants apply only to the linked organization, used through one of its project API keys; changing the linked organization through support; plan changes, Plus, Go, Free and Enterprise ineligible
- Changelog | OpenAI API — October 28, 2026: eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in free API credits as a subscription benefit
- Spend limits | OpenAI API — Hard limits return 429 with organization_spend_limit_exceeded or project_spend_limit_exceeded; check error.code to tell those, organization_usage_limit_exceeded and credit_balance_exhausted apart; for credit_balance_exhausted, add credits; a project hard limit applies only to traffic billed to that project
- Integrations OpenAI — Fours Doc — Org-level connection uses an OpenAI API key for org-wide access, billed to the organization and handled directly by OpenAI; optional Organization ID for someone in multiple organizations; the API key can't be edited, so changing it means deleting the integration and creating a new one; the user-level connection is a Codex sign-in authorized at OpenAI
- Insulin Agents — Fours Doc — Model Selection: a failed model call retries on the next connected provider, Fours-hosted models the fallback; the built-in Insulin assistant and personal agents add the providers each member connected and never reach the organization's keys, organization agents add the organization's; Default model set on a custom agent; Slack and Microsoft Teams reach the built-in assistant; When a model fails over: a short italic note naming both models
- Insulin Getting Started — Fours Doc — Adoption Path: organization integrations connected or disconnected only by an org ADMIN; AI Gateway: a non-empty allow-list limits AI features to the integrations on it and drops the hosted pool unless Fours Hosted (DeepInfra) is listed, and Allow Fours platform key off makes the organization bring-your-own-key only; When a connection expires: API-key connections such as OpenAI never show the expired state
- Insulin What Is Metered — Fours Doc — Using your own provider keys: the provider bills the tokens under your own contract and Fours charges a flat fee per unit; a zero balance still pauses BYOK requests because the per-unit fee is still owed
- Insulin When AI Pauses — Fours Doc — On pay as you go, AI features pause when the balance reaches zero; monthly billing is never stopped for a balance
- Insulin Billing FAQ — Fours Doc — Monthly free credit expires at the end of the month it was issued in; purchased credit stays valid for three years
Keep reading
Stay Updated
New posts, product updates and marketplace strategy are shared on LinkedIn as they publish.
Follow Fours on LinkedIn