OpenAI’s API changelog entry for October 28 says eligible ChatGPT Pro and Business subscribers “can now claim up to $500 per month in free API credits as a subscription benefit”, activated by linking an API organization the subscriber owns. The changelog shows a date with no timestamp, and the guide it points to carries no date, so October 28 rests on the changelog alone. The guide’s table gives a Pro grant of $100, $200 or $500 a month, matching the tier prices. Business workspaces get $20 per paid Standard seat or $100 per paid Premium seat, pooled and capped at $500 a month. API changelog · API credits guide
S5 Labs has not claimed the benefit or run traffic against it. Everything below comes from OpenAI’s developer documentation, read on October 28. openai.com and help.openai.com refuse scripted fetches, so the service-credit terms the guide links to and the help-center pages on Pro tiers and Auto Recharge were not read.
Who gets how much
| ChatGPT subscription | Monthly API credits |
|---|---|
| Pro ($100 tier) | $100 |
| Pro ($200 tier) | $200 |
| Pro ($500 tier) | $500 |
| Business Standard | $20 per paid Standard seat |
| Business Premium | $100 per paid Premium seat |
The first grant is prorated for the time left in the current billing cycle and expires with it. Unused credits “expire at the end of the billing cycle and don’t roll over.” A Business workspace that pays annually gets the year’s grant in one issuance, capped at $6,000, expiring at the end of the annual cycle. Adding seats or upgrading seat types mid-cycle can add prorated credits under the cap; seat reductions take effect next cycle; reassigning a seat to another person creates nothing. API credits guide
The seat arithmetic is ours: 25 Standard or 5 Premium seats reach the cap, 10 Standard seats earn $200, 3 Premium seats earn $300. The three Pro tiers are the ones from DevDay, covered in our DevDay article.
Eligibility and claiming
The subscription must be active, paid and at least seven days old; trials and complimentary plans do not count. One person must own both the ChatGPT subscription or Business workspace and the receiving API organization, using the same email address on both accounts. Someone new to the API can link their Personal API Organization. API credits guide
Two claim paths make the same link. In ChatGPT, Pro users go to Settings, then Billing; Business owners use workspace billing settings or the Admin Console. On the API Platform, pick the receiving organization, open Billing, find the benefit and select the subscription to link. Credits then show on the billing overview. The guide says credits arrive “automatically when your subscription renews”, and does not say whether the first grant posts at claim time or at the next renewal. Changing the linked organization later means contacting support. API credits guide
What is covered and what is not
The credits cover “paid API Platform usage billed to the linked organization”. The guide lists the Responses API, the Agents API “including OpenAI-hosted sandbox charges”, the Decisions API, GPT-Live and the Realtime API, the Images API, and built-in tools with their usage and storage charges, naming web search, file search and Code Interpreter. It adds that other APIs are available through the platform, without saying every model or endpoint qualifies. API credits guide · Our Agents API coverage · Our Decisions API coverage
Two exclusions: “Codex usage through API keys is not supported by this benefit”, and the credits “don’t increase your ChatGPT or signed-in Codex usage limits.” A Pro subscriber running Codex on an API key keeps paying for it. On plans, Plus, Go and Free are named as ineligible, and Enterprise “is not eligible for this benefit”; Edu is not mentioned. Cancelling stops new grants without revoking credits already issued. A downgrade to an ineligible plan, or a Business move to Enterprise, removes the organization link at the end of the cycle. Moving from Pro to Business ends the Pro link; the workspace owner can make a Business link once the paid workspace is seven days old. API credits guide
The 429 when the balance hits zero
Free credits are spent before purchased ones. When the free balance is gone:
| Organization balance | What happens |
|---|---|
| Purchased API credits remain | API usage continues using your purchased credits. |
| No credits remain | API requests fail with HTTP 429 and error code credit_balance_exhausted. |
The guide says no API payment method is required, so an organization can run on grant credits alone, and that organization starts returning 429s the moment the grant is spent, which on a $100 plan may be well before the cycle ends. The fixes offered are adding paid credits or setting up Auto Recharge “to add paid credits to continue building without errors and interruptions”. Auto Recharge’s minimums and failure behavior sit in a help-center article we could not read. API credits guide
The error-codes page lists credit_balance_exhausted as a 429 whose cause is that the organization “has no prepaid credits remaining”. It lists other 429s too: the generic rate limit with a Retry-After header, slow_down, organization_spend_limit_exceeded, project_spend_limit_exceeded and organization_usage_limit_exceeded. Our inference: a handler that backs off on every 429 will loop on the credit and limit codes, which clear only when someone changes billing or a limit, so branch on the code and alert on those. The guide never says whether the grant counts as a “prepaid credit” in the error page’s wording, though it states the 429 outcome directly. Error codes · API credits guide
How far $200 goes
Standard rates per million tokens, for prompts up to 272,000 input tokens:
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6.1 Terra | $1.00 | $0.05 | $5.00 |
| GPT-6.1 Sol | $2.00 | $0.10 | $10.00 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 |
Pricing · Our GPT-6.1 Terra article
Our assumptions: reasoning tokens bill as output, no tool, sandbox or storage charges, cache-write fees ignored, no regional or FedRAMP uplift, no request above 272,000 input tokens. Spent on a single model and token type, $200 buys 200 million input tokens or 40 million output tokens on Terra, 100 million or 20 million on Sol, 2 billion or 400 million on Luna. At a 3:1 input-to-output mix the blended price is $2.00 per million on Terra, $4.00 on Sol and $0.20 on Luna.
| 20K input / 2K output | Per request | Requests for $200 | 16K of input cached | Requests for $200 |
|---|---|---|---|---|
| GPT-6.1 Terra | $0.03 | 6,667 | $0.0148 | 13,514 |
| GPT-6.1 Sol | $0.06 | 3,333 | $0.0296 | 6,757 |
| GPT-6 Luna | $0.003 | 66,667 | $0.00156 | 128,205 |
Multiply by 2.5 for the $500 tier and halve for $100. Batch and Flex cut Terra to $0.50 input, $0.025 cached and $2.50 output, so the uncached request costs $0.015 and $200 covers about 13,333 of them. Fast mode doubles Terra to Sol’s Standard price. These are token budgets, not finished tasks; an agent that retries or reasons at length gets through fewer. Pricing · GPT-6.1 Terra model page
The usage-tier question OpenAI has not answered
The benefit is subject to API rate limits. The rate-limits guide lists a Free tier with a $100 a month usage limit, then Build, Launch and Grow at $5, $100 and $500 “in total credit purchases”, with monthly usage limits of $500, $5,000 and $200,000. OpenAI also sets an approved monthly usage limit per organization, separate from the spend limits you configure. Neither page says whether grant credits count as purchases, or which usage limit applies to an organization holding only grant credits. A $200 or $500 grant sits above the Free tier’s $100 a month, so an organization that never bought credits might be unable to spend its whole grant. We are raising the question, not asserting the ceiling; the organization limits page is where to check. Rate limits · API credits guide
What to check
Start with ownership. If a company card pays for Business and the API organization belongs to an engineer’s account, the link fails until one person owns both on one email.
Then pick the organization to link with care, because credits land only there and moving them is a support ticket. Link the organization whose project keys the agent uses.
Before scheduling anything unattended against a grant, confirm a paid balance or Auto Recharge. Then force the failure once and make sure your client pages someone on credit_balance_exhausted instead of retrying.
Codex sessions on API keys keep billing to the paid balance.
Read the organization’s tier and approved monthly usage limit. If it is below the grant, that is the real ceiling until OpenAI says otherwise.
The guide links its “API terms” to OpenAI’s service-credit terms, which we could not read. The guide itself states no transfer or refund rules, so check those terms before counting the grant in a budget.
Sources
- API credits for ChatGPT subscribers — OpenAI Developers
- API changelog — OpenAI Developers
- Error codes — OpenAI Developers
- Rate limits — OpenAI Developers
- Pricing — OpenAI Developers
- GPT-6.1 Terra model page — OpenAI Developers
- Service credit terms — OpenAI (linked from the guide, not read: 403 to scripted fetch on October 28)
- GPT-6.1 Terra pricing — S5 Labs
- Decisions API — S5 Labs
- Agents API — S5 Labs
- DevDay 2026 — S5 Labs