Skip to content
API referenceGLM: errors and limits

GLM: errors and limits

Exact-two catalog, queue behavior, Responses continuation, and honest unknown usage.

Public generation accepts only glm-5.3 and glm-5.3-flash. Short aliases, the previous seven GLM IDs, Turbo, and FlashX return 404 model_not_found before reserve or debit.

ResultWhat to check
400 invalid_requestCheck the selected Messages, Chat, or Responses body and parameter compatibility. Unsupported combinations are rejected before money.
401 invalid_api_keyUse the complete active gd-… key in the correct header.
402 insufficient_balanceCheck available balance and unfinished reserves in the dashboard.
404 model_not_foundUse one of the two exact model IDs without an alias.
429 or waitingThe pool uses a fair queue for up to 10 seconds. Retry after the stated delay; do not hide parallel retries inside a stream.
Unavailable previous_response_idContinuation is pinned to the previous account and same key; state is not moved to a fallback account.

Documentation sections

On this page