Run ComfyUI workflows or deploy open models on an on-demand GPU (RTX 5090, 4090 and A100). CloudGPU offers an OpenAI-compatible API and an MCP server, plus a hosted-model API with about 50 models. Billing comes from a prepaid balance, per minute on some configurations and per hour on others, and the quote shows the billing period and minimum charge before you start. This is a public beta with no SLA. CloudGPU is an independent service and not the official Claude API.
CloudGPU addresses a common practical problem: a short AI job can require more GPU memory than a developer has locally. We give teams a way to choose a configuration and review a quote before starting.