GPU power as an API
Buy prepaid credits and run real GPU workloads across the whole RentGPUPower cluster from any script — no browsing, no fixed terms. Serverless jobs, or grab a GPU by the second.
Built for fleets: ask /capacity before you fan out, queue in line when the pool is full, get a signed webhook the moment a job finishes, hand teammates read-only keys, and import the OpenAPI spec into your tooling.
Which API do I want?
Everyone's compute, on demand
The pool API: serverless jobs and OpenAI-compatible chat that run on whichever verified machine is cheapest and idle right now — the general compute of the whole marketplace. Pay per second from credits; wait in line with queue_s when it's full, and be told by webhook when you're done.
Jobs & chat endpoints →One specific machine, by the second
Instances: reserve exactly the hardware you want — a model, a rig with min_gpu_count, a region — and hold it as long as you pay. You get that machine's own gpurent_… key.
Instance endpoints →A rental you made on the website
Every marketplace rental — picked by hand from a listing — has the same per-rental API: run commands, move files, open web services on that one machine with its gpurent_… key.
Per-rental API docs →