Build, deploy, and operate GPU workloads.
Move from a first GPU deployment to repeatable AI infrastructure with guides for compute, SSH, APIs, serverless inference, billing, security, and support.
Guides / API / Examplescustomer-visible workflows26 published guides · platform alignedChoose a workflow and its completion check
For a custom environment, begin with GPU cloud operations: review an exact offering and authorization, wait for readiness, verify SSH identity and run a small application test. Export outputs and confirm termination when finished. Closing the terminal is not the completion check.
For a hosted model request, use the serverless quickstart: discover the public model ID, validate the supported schema, authorize one bounded request and inspect output and billing. Current chat returns completed JSON rather than token streaming. Reconcile an uncertain timeout before retrying.
For automation, the API reference defines scopes, fields and response envelopes. Keep model discovery separate from billable generation and quoting separate from launching. If an operation fails, retain UTC time, safe request or instance identifiers and the first redacted error for support; never include credentials.