OpenAI-compatible API with a per-workload key — drop-in for existing SDKs.Dedicated GPUs — your model, your instance, no shared queues.Billed by the second with auto-shutdown — never pay for idle time you didn't plan.