Skip to main content

Insufficient funds

Prime Inference returns this error when the account billed for the request does not have enough balance. The API key identifies the key owner’s personal account, and the X-Prime-Team-ID header switches billing to that team. The Prime CLI team setting does not apply to direct API requests.
Requests without X-Prime-Team-ID bill the API key owner’s personal account.
  • Add funds to your personal balance in the billing dashboard
  • Check that the API key belongs to the account you funded
  • Retry the request
Requests bill a team only when they include X-Prime-Team-ID, and the key owner must have access to that team.
  • Add funds to the team’s billing balance
  • Include X-Prime-Team-ID on every request, including SDK requests (see team billing)
  • Find the team ID on the team profile page or with prime teams list

Deployment command unavailable

If prime inference deploy or deploy-checkpoint is missing, update to Prime CLI 0.21.0 or later:
For dedicated endpoints, use prime inference deploy or prime inference deploy-checkpoint. See the deployment CLI reference for the supported command names and options.

Dedicated endpoint still deploying

Inspect the existing deployment:
Check status, statusPhase, and errorMessage, or open dashboardUrl for progress. Available GPU capacity does not mean weights have loaded. Startup time varies by model and whether weights are cached. If --wait times out or loses contact with the API, the endpoint can continue starting. Check its status before retrying creation. Delete it if you no longer need it, because it can start serving and incur charges after your command exits. If the deployment reports an error, include its ID and error message when you contact support. Avoid creating repeated deployments to diagnose the same failure.

Team deployment missing from the CLI list

Pass the owning team explicitly:
In Prime CLI 0.21.0, omitting --team lists personal deployments even if prime config set-team-id selected a team. Pass --team explicitly to list a team’s deployments. For a deployment ID you already have, prime inference deployments get <deployment-id> checks access directly.

Dedicated endpoint requests fail

  1. Confirm that the deployment is RUNNING and has a non-null endpointUrl.
  2. Copy servedModelAlias exactly into the request’s model field. The catalog ID, deployment name, and base model repository do not select the endpoint.
  3. Use an API key with Inference permission and access to the endpoint’s account.
  4. For a team endpoint, add X-Prime-Team-ID to every direct HTTP or SDK request. For a personal endpoint, use the owner’s key and omit that header.
A 404 model_not_found can mean the alias is wrong, the deployment has not reached RUNNING, or it has been deleted. A 403 forbidden can mean the request uses the wrong account context; check the API key’s access and the X-Prime-Team-ID header. CLI model requests use the selected team from prime config set-team-id. That setting does not add headers to requests made by your own HTTP client or SDK. See Send a request for examples.

Checkpoint unavailable

List the run’s saved checkpoints:
Pass a listed step explicitly to prime inference deploy-checkpoint <run-id> --step <step>. For runs that save to a volume, wait until the run is completed, stopped, or paused and the final volume sync has finished. Only complete checkpoints on a finalized volume are available; checkpoints still on an active run’s volume cannot be deployed through this workflow. If the base model has no serving configuration in the inference catalog, contact support. If the run’s own cluster cannot serve the model, use on-demand capacity through the dashboard’s Custom weights placement step or the checkpoint API.

Replica change rejected or still in progress

Only RUNNING deployments can be scaled. Inspect maxReplicas, gpusPerReplica, scaleState, and errorMessage with prime inference deployments get <deployment-id> -o json. A successful scale request records the desired count before workers are ready. On-demand scale-ups also need available pool GPUs and sufficient wallet balance within account limits. If the returned state remains blocked or reports an error, check the endpoint page before submitting another scale request. To stop the endpoint entirely, delete it instead of requesting zero replicas.

Getting help

If your issue isn’t covered here: