Why Free Token Credits Can Still Leave You at a Usage Limit
The quick answer
You can have free credits in your account and still hit a usage limit. The credits may cover a different model or API route, while the session you're running uses the normal subscription allowance.
Build your budget shortlistIn this guide
You can have free credits in your account and still hit a usage limit. The credits may cover a different model or API route, while the session you're running uses the normal subscription allowance. Before claiming more, check which balance this request is supposed to use.
Our deals and free-offer directory can help you check recorded offer conditions, while your account remains the place to confirm whether a benefit is active.
A large promotional balance makes the error feel contradictory: why is the tool asking you to wait when the account still shows plenty left? Follow the request from the app to the model and billing pool. Somewhere along that path, the offer's conditions may differ from the way you're using it.
The way to resolve it is to follow one request from client to entitlement.
Write down what the offer actually grants
Open the campaign page or the offer shown in your account. Record its exact name, eligible model, permitted product, activation requirement, expiry date, and any rate limits. If the page simply says “tokens,” look for whether input, cached input and output are treated differently.
Do not fill missing conditions with assumptions from another promotion. A previous offer for the same provider may have used different models or a different application. A promotional API balance also may not appear inside a consumer chat subscription.
Use this small checklist:
| Question | Evidence you need |
|---|---|
| Is the offer active? | Account status or activation confirmation |
| Which model is eligible? | Campaign terms and selected model |
| Which client or endpoint is eligible? | Official integration instructions |
| Which balance should decrease? | Usage dashboard showing the relevant pool |
| When does access end? | Expiry and remaining quota, checked separately |
The expiry date answers how long you can claim or use the benefit. It does not tell you how quickly you can spend it.
Check the actual route
The client name is insufficient. An agent can use a provider subscription, a direct API key, or a gateway. Inspect the sign-in method and base URL without copying secrets into screenshots or support messages.
For GLM, the official integration guide documents different protocol endpoints and says the wrong endpoint can prevent subscription-quota use. Follow the guide for the client you are actually running. Do not substitute a generic endpoint solely because a request format looks compatible.
If the interface exposes a model alias, check what that alias currently maps to. An interface label such as “fast” does not prove that the request uses the model covered by your promotion. Avoid repeatedly changing several settings at once; otherwise you cannot identify what fixed the problem.
Separate balance, window and rate
Three independent conditions can apply to the same request. You may have a positive credit balance, have exhausted a five-hour allowance, and be below a per-minute request limit. Passing one condition does not imply passing the others.
The GLM plan overview currently describes both five-hour and weekly credit windows, with model usage weighted by token type. Check promotional terms separately from these base-plan rules. The plan FAQ also describes campaign reset cards separately. Do not treat a token grant and a reset card as interchangeable benefits.
A useful diagnosis is specific: “The promotional pool is visible, but this selected model is deducting from the ordinary plan.” “The app is broken” is too broad to guide the next step.
Run one small, observable check
Once the documented configuration is in place, note the relevant balances and send one short request using non-sensitive sample content. Check the response's model information if available, then inspect the usage entry after any documented reporting delay.
You are looking for a matching record, not an exact deduction guessed from the length of the displayed answer. Hidden context, tool results and billing weights can affect usage. If the dashboard does not expose enough information, record that uncertainty instead of claiming the promotion is working.
If the request fails, preserve the error category, time and timezone. Stop repeating it when the error clearly indicates a reset window. Repeated attempts will not establish eligibility and can make the evidence harder to interpret.
Escalate with the missing link
If the offer is active, the route matches, and usage still comes from the wrong pool, contact the provider with those facts. Include the campaign name, selected model, client version and a redacted error or usage record. Never send an API key.
If the offer is active and the route matches its terms, ask support to explain the specific deduction or limit you're seeing. A campaign name, model ID and redacted usage record are more useful than a screenshot of the headline balance. More credits won't help until you know where the existing ones can be spent.