Launched October 27, 2026
lium.io

lium.io

Agent-first GPU cloud for renters and providers

One time feeAICodeHosting767 impressions#29 of its week2 comments

Comments

>log in to comment
  • Fatisco Stack Industries· 9d ago

    When a rented GPU pod gets terminated mid-job, is the per-second billing cut off at that exact moment, or does it run until the provider releases the node?

    • lium support[maker]· 7d ago

      Billing stops when the pod ends. It never waits for the provider to release the machine. - If you delete the pod, charging stops at your delete request. - If the pod breaks because its node fails, billing stops then. On Secure nodes you are also credited for the interrupted rental. - If the node stops responding while your pod runs, billing pauses until the node is back. Outage time before the pause is refunded automatically. Time in which your pod still answered SSH stays charged.

Agent-first GPU cloud. Rent NVIDIA GPUs via web, CLI, SDK, or API. Providers earn 95% rental revenue + bonuses. Usage-based GPU rentals. Documentation: https://docs.lium.io. Providers: https://provider.lium.io. CLI/SDK: https://github.com/Datura-ai/lium.

More AI tools →74 free dev tools →

lium.io provides on-demand GPU rentals via web, CLI, SDK or API with per-second billing.

for
Developers and AI agents needing scalable GPU compute on demand.
pricing
paid
license
MIT
Datura-ai/lium 7 3Pythonupdated a month ago
works withCLISDK

Key features

5 features of lium.io
  • Per-second billing — Pay only for the exact seconds a GPU pod runs, with no minimum reservation.
  • Wide GPU catalog — Choose from 92 GPU models ranging from consumer RTX cards to data-center H100s.
  • CLI/SDK/API access — Spin up, manage and terminate pods programmatically via pip install, CLI commands or SDK.
  • Transparent pricing API — Access live pricing and node lists as JSON without an account.
  • Provider revenue share — Providers keep 95 % of rental revenue plus bonuses.

Use cases

  • Training large AI models on high-end GPUs without owning hardware.
  • Running inference workloads for AI agents that need burst GPU power.
  • Scaling GPU resources for CI/CD pipelines that include GPU-accelerated tests.
  • Providing on-demand GPU compute for research experiments with variable workloads.

lium.io FAQ

What do the prices on the pricing page represent?+

Each figure is the hourly price for one GPU set by the pod’s provider; ranges show live asks across available pods.

Does Lium have per-second billing with no minimum?+

Yes, pods are billed per second with no minimum, though you need a balance covering 15 minutes to start a pod.

How can a script read current prices?+

The pricing table is available as JSON at lium.io/pricing.json and node data at lium.io/api/public/v1/nodes without authentication.

Are storage or platform fees included in the hourly price?+

The hourly price includes local disk; persistent volumes are billed separately per GB-hour plus data egress.

Edited by the makers, based on lium.io · Oct 15, 2026. Details may change; check the official site.