This task can be performed using Lium.io
Agent-first GPU cloud for renters and providers
Rent GPUs from code
For cloud GPU rental, first create an account, configure CLI or API credentials, and confirm billing limits. Decide the NVIDIA model, GPU count, region, image, storage, and maximum hourly cost before submitting a rental request.
Use the CLI for interactive setup and repeatable shell scripts. Use the API for schedulers or agents: query availability, create the instance, poll until ready, connect, run a small CUDA check, then stop and delete resources when finished.
Best product for this task
Lium.io
cloud
Agent-first GPU cloud. Rent NVIDIA GPUs via web, CLI, SDK, or API. Providers earn 95% rental revenue + bonuses.
What to expect from an ideal product
- CLI and API feature parity
- Clear GPU availability and pricing
- Flexible GPU and image selection
- Reliable provisioning status and errors
- Simple stop and deletion controls
More about safe automation
API requests can fail after submission even when a GPU was allocated. Treat creation as a stateful operation, and verify the resulting instance before retrying.
- For automated GPU workflows, record a request ID and check existing rentals before creating another.
- Set explicit timeouts for provisioning and handle failed or unavailable states.
- Run nvidia-smi and a minimal CUDA test before transferring data.
