Get model status
GET
/v1/model-status
Poll this to see whether the GPU model backend is up and able to process pages right now. state is one of ready, starting, or unconfigured — while it's starting, wait and poll again rather than kicking off a batch that would fail.
Responses
OK
application/json
Success body for GET /model-status.
The configured model backend's name.
A short, customer-safe description of the state — never a device name, endpoint
id, or other internal detail.
Whether real models run (RunPod) rather than none being configured.
True when a translation request would succeed right now.
One of ready, starting, throttled, or unconfigured.
The coarse cold-start progress while state is starting — one of
initializing / loading_models, or null when the backend is not warming
(#2149). Lets the readiness banner tell the user whether the worker is spinning up or the models are
loading during the multi-minute cold start.