Renter FAQ
Short answers to the questions renters ask most. Each answer links to the page with the full detail.
Is Lium reliable?
Lium checks the machines it lists and lets you pick nodes by their track record. The Lium team's validator checks every connected node every 15 minutes, and Browse Pods filters nodes by the provider's reliability score over recent rentals, by uptime and by download speed. Each node is listed as Secure (dedicated) or Spot (may be reclaimed). See Validators and Broken pods.
What happens if my pod or its node fails?
You are not charged for a pod that fails to start, and on Secure nodes you are credited when a provider force-closes your pod. If a pod fails to start because of its node, Lium starts it on another node with the same GPU at the same or a lower price; if no such node is free, the pod ends CREATION_FAILED and is not billed. A pod the provider force-closed shows BROKEN: billing has stopped, and on a Secure node you have been credited. If a node stops reporting, its pods are charged $0 once it is marked inactive (after an hour without a validator report), and what they were charged between its last report and that mark is credited back. See Broken pods and If your pod disappears.
What if the provider takes the machine down?
The provider must announce it ahead of time: the pod page shows a notice, and you get an email if your pod was already rented. Scheduled maintenance lasts at most 60 minutes and is announced at least 24 hours ahead; afterwards Lium restarts your pod, and files in its local volume and on an attached Volume stay. A permanent removal is announced at least 48 hours ahead; the pod is then deleted, while your Volumes and completed backups stay in your account. See Provider maintenance on your pod.
Is my data secure on Lium?
It depends on the machine you rent. On an ordinary (non-CVM) machine the GPU provider has host-level access and can read your pod's files, memory and environment variables, so never put private keys, seed phrases or API keys on it; /root is encrypted by default on images that support it, which protects data at rest, not a running pod. Confidential Computing (CVM) machines use Intel TDX hardware isolation, so the provider cannot read the VM's memory or disk. Accessing a renter's pod is forbidden for providers on any machine; report a suspected case through official Discord support. See Pod security and Encrypted local volumes.
Can I run a private model the provider cannot see?
Yes, on a Confidential Computing machine with a key service you control. Your pod proves from the inside what machine it is on with a hardware-signed quote, your verifier checks it, and only then releases a short-lived key that decrypts your model or dataset, so nothing secret is stored in the image, the template or your Lium account. This keeps the GPU provider out, not Lium, whose node software runs inside the VM. See Private models and datasets.
How much does it cost?
You pay the hourly USD price per GPU that the provider sets for the machine, times the number of GPUs you rent. Live prices for every GPU model are at lium.io/pricing, and lium ls lists the nodes you can rent now with their prices. Volumes and backups cost extra, per GB of data stored per hour. See Billing.
How is billing done?
Per second, from your Lium balance. A pod is charged its hourly price × seconds / 3600 for the exact seconds it existed, counted from the deploy request once it reaches RUNNING, so provisioning is included. Your balance is charged every 5 minutes and once more when the pod is removed, and a pod that never starts is not charged. See Billing.
Is there a minimum rental time?
No. There is no minimum and no rounding up to a minute or an hour: a pod at $2.40/hour deleted after 20 seconds costs $0.0133. To deploy, your balance must cover at least 15 minutes of the pod's hourly price. See Billing.
How can I pay?
By topping up your balance with a card, stablecoins, TAO or alpha. Card and stablecoin top-ups start at $10, and Auto Top Up can refill your balance from a saved card when it runs low. Card top-ups come with an invoice that can carry your company name, address and tax ID. See Billing and Invoices.
Which GPUs can I rent?
Whatever the providers list at the moment, for example the RTX 3090, RTX 4090, A6000 and H100. Every GPU model with its live price range and availability is at lium.io/pricing, and lium ls --gpu <model> lists the nodes you can rent now. See Your first hour.
How fast does a pod start?
Usually in under a minute: 20–60 seconds is typical with the default image, and our timed run was ready in 42 seconds. A node that has to pull the image first can take several minutes. See Your first hour.
How do I keep my data when a pod ends?
Put it on a Volume or in a backup: anything else on the pod is gone when the pod is removed. A Volume is persistent storage mounted at /mnt that survives the pod and can be attached to another pod; backups upload a directory of the pod's local volume to S3 on a schedule, and you can restore one into a new pod. Push finished results to storage you control as you go, too. See Volumes, Backups and If your pod disappears.
Can I use Lium from the CLI, an SDK, the API or an AI agent?
Yes. One package, lium.io (pip install lium.io), ships the lium CLI and the Python SDK, and the REST API takes an API key in the X-API-Key header and publishes a live OpenAPI spec. For AI agents there is a non-interactive CLI contract, a docs MCP server and llms.txt. See Developers and AI agents.
How do I get support?
Through the official Lium Discord: open a private ticket from the #tickets channel, or click Request Discord Support on a rented pod's page for a private channel with the provider and Lium support. Never share passwords, seed phrases, private keys or API keys with anyone who contacts you, even if they look like Lium staff. See Discord support and Official support and scam prevention.