Fix 503 model_unavailable and "No available channels"

What 503 model_unavailable, "No available channels" and overloaded_error mean on BigModel.io, how long they last, and how to fall back to another model automatically.

On this page
  1. What it means
  2. How to fix it
  3. Troubleshooting
  4. Related

The model exists, but it cannot take requests right now:

json
{ "code": "model_unavailable", "message": "The model `X` is temporarily unavailable. Please retry shortly.", "error": { "message": "The model `X` is temporarily unavailable. Please retry shortly.", "type": "api_error", "code": "model_unavailable" } }

#What it means

Message Cause Retry?
model_unavailable "temporarily unavailable" The model has no capacity at the moment Yes, after a short wait
api_error "No available channels" The model is briefly not on sale Later, or use another model
overloaded_error The model service is overloaded Yes, with backoff
billing_service_error The billing check failed Yes

None of these are charged.

#How to fix it

  1. Retry with backoff, starting at a few seconds. Most cases clear within minutes.
  2. Fall back to another model. Send a list of models; the first one that is available answers:
json
{ "model": "glm-5.3", "models": ["glm-5.3", "deepseek-v4-pro", "kimi-k3"], "messages": [{ "role": "user", "content": "Hello" }] }

The fallback list is used when a model is unknown, retired or unavailable. It does not retry after a 429 or 502.

  1. Check the model catalog for a model with the same endpoint and similar price.
  2. If one model stays unavailable for a long time, contact Support with the X-Request-ID.

#Troubleshooting

Only streaming requests fail. A stream that fails after it starts ends with an error event instead of a 503. See Upstream request failed.

An image or video task stays queued. Media tasks run asynchronously. Poll the task instead of retrying the create call. See Media tasks.

Updated · View as Markdown