Lium

Lium rents GPU pods such as H100, H200, B200 and RTX 4090 from independent providers at marketplace prices, billed by the second. It targets developers and AI agents that need compute for training, inference or short jobs, and pods start from the browser or a CLI.

Each listing shows the GPU model, CPU, memory, disk, network speed, uptime and the provider's location, so you can choose hardware by price and reliability. Spot pods sit alongside regular ones, and templates, storage volumes, backups and SSH or Docker credentials are managed from the same dashboard.

For automation, a Python SDK, REST API and CLI let an agent start a pod, run a job and tear it down within a budget and time limit you set. Anyone with spare GPUs can also sign up as a provider and list hardware on the marketplace.