Deployments
The Deployments section manages your live inference endpoints. It mirrors client.deployments in the SDK. Each endpoint serves a single model behind an HTTPS API.
Create a deployment
From the Models section, click Deploy on a ready model. Choose the hardware tier and Fleet provisions a serverless endpoint. Inference scales up with demand and back down to zero when idle, so an idle endpoint costs you nothing. See Billing for how inference is metered.
Deployment status
| Status | Meaning |
|---|---|
starting |
The endpoint is being provisioned. |
active |
Live and serving requests. |
failed |
Provisioning or startup failed. |
stopped |
Shut down; no longer serving. |
Test inference
Click Test (or open the endpoint detail) to send a prompt straight from the dashboard and see the response, without writing any code. This is the quickest way to confirm a freshly deployed model behaves as expected.
For production use, copy the endpoint's URL and call it from your app, or use endpoint.infer(...) in the SDK. Responses can be streamed.
Manage deployments
Each endpoint row offers actions to copy its URL, test it, and delete it. Deleting an endpoint stops it serving and releases its resources immediately.