API Key Limits & Credit Billing Rules

One shared account balance ($1 = 100 Credits) supports unlimited server-side API keys with optional daily caps, reset timezones, model allowlists, and post-completion billing.

$1 USD = 100 Credits · Min Top-Up $10

Server-side Bearer keys & per-key limits

Pass Authorization: Bearer cm_live_... from backend services only — never expose API keys in browser or mobile client code. All keys share your account credit balance while enforcing optional per-key limits:

ControlConfigurationRuntime Behavior
Daily Credit LimitOptional credit cap + IANA timezone (e.g. Asia/Shanghai, UTC)Counts settled charges + active holds in the current local day; blocks excess with 402 key_quota_exceeded.
Total Credit LimitOptional cumulative cap per keyCaps cumulative spend across the key’s lifetime — ideal for client projects or test budgets.
Model AllowlistAll models or selected public model IDsUnselected models are rejected immediately with 403 model_not_allowed (0 Credits charged).
Expiration & RevokeOptional expiry date or instant disable/revokeTakes effect immediately for new requests without affecting other keys in your account.

Fixed credit rate & 3 billing modes

Callmodels uses a fixed rate of $1.00 USD = 100 Credits (minimum top-up $10.00 = 1,000 Credits). Every accepted request locks the published model price at submission, temporarily holds the specification ceiling, and deducts only when your output file is ready:

Zero-charge failure guarantee: Whenever a task ends in failed (invalid input, content policy block, generation error, or timeout), 100% of the held credits return to your account balance and API key quota.
Billing ModeModelsInitial Hold (HTTP 202)Final Charge on Completion
Actual tokens
gpt-image-2.6-flaregpt-image-2.6-sunburst
Standard output ceiling + prompt/reference token bufferActual metered tokens × published per-1K rate; unused hold returned immediately.
Fixed per imageseedream-5.5Fixed per-image rate (1.88 or 3.29 + 0.752 × refs)Exact same amount (0 variance across aspect ratios or prompt lengths).
0.1s durationSeedance 2.5 / Flash Seedance 2.0 / MiniRequested duration × rate_per_secondActual output video duration rounded up to 0.1s (4.0s minimum, capped at requested duration); difference returned.