A rental gives you a real GPU, a browser notebook and a root terminal — so anything that runs on an NVIDIA card runs here. Pick a workload and launch on an available GPU in minutes.
Click any workload to see the GPUs available right now, pick one, and your instance starts — ready for that job in about a minute.
What we don’t do (yet)
We’d rather tell you the limits up front than have you discover them mid-project:
No public API endpoints.You can run a server like vLLM inside your instance and call it from your own notebook or scripts in that session, but we don’t give you a public URL for the outside world to hit.
Not for always-on production. Rentals are interactive and per-second; a host can go offline. Use us for training, experiments, batch jobs and prototyping — not for the service your customers depend on.
Single machine, not a cluster.You get every GPU on one host, but we don’t link several hosts into one distributed training run.
Availability varies.We’re a young marketplace, so the exact card you want may not be online at that moment. See what’s available now.