Check access before you troubleshoot

Platform reachability

The public website and status API are monitored as separate checks from individual model requests. A reachable platform does not guarantee that every upstream model is available.

For live model availability, open the model catalog and retry only with bounded backoff. Preserve the request identifier from an error response when contacting support.

Model-specific incidents

Availability can differ by model, modality, or capacity. If a model fails repeatedly, verify the exact model identifier and request format, then test an appropriate alternative from the catalog.

No public SLA percentage is promised on this page. Contractual support or availability terms apply only when separately agreed in writing.

Get help

Email [email protected] with the approximate time, request identifier, endpoint, model name, and HTTP status. Never email an API key or full prompt containing sensitive data.