GPU Router
Loading models…

Fetching the model catalog.

DEVELOPER DOCUMENTATION

Errors & limits

Handle errors deliberately. Keep retries bounded.

HTTPMeaningNext step
400Invalid requestCheck the exact model ID, text messages, and supported parameters.
401Invalid or revoked keyCheck your key and selected gateway.
402Insufficient creditsCheck active stake and available credits.
404No available routeRefresh the live catalog; the model may no longer be available.
429Rate limit reachedBack off before retrying.
502 / 503Upstream or service unavailableRetry with bounded exponential backoff; inspect any receipt first.

Request limits

Maximum output is 65,536 tokens at the gateway and may be capped further by available credits or upstream limits. The model’s catalog context and output limits do not override the gateway’s request contract. Owner-level rate limits are shared across that owner’s keys.

Documentation | GPU Router