lium ls
List available GPU nodes.
lium ls [OPTIONS]
Options
| Flag | Effect |
|---|---|
--gpu TYPE | Filter by GPU type (H100, A100, RTX4090, …) |
--count N | Exact GPU count to match (e.g. 1, 8) |
--min-cuda VERSION | Minimum CUDA version supported by the node (e.g. 12.4) |
--country CODE|NAME | Only nodes in these countries — ISO code (exact) or country name (case-insensitive prefix); repeatable or comma-separated (--country US,NL) (since 0.0.40) |
--min-vram GB | Minimum memory per GPU, compared on the same rounded GB the table prints (--min-vram 80 keeps an 81559 MiB H100) (since 0.0.40) |
--max-price USD | Maximum price per GPU-hour (2.50) (since 0.0.40) |
--tier spot|secure | Only spot (reclaimable) or secure nodes — see the Tier column below (since 0.0.40) |
--min-cpus N | Only nodes that report at least N CPU threads — the CPUs column (cpu_count in JSON; — when the node does not report it) (since 0.0.40) |
--lat LAT | Latitude for distance filtering |
--lon LON | Longitude for distance filtering |
--max-distance MILES | Maximum distance in miles from --lat/--lon |
--sort FIELD | Sort by price_gpu (the default since 0.0.40; download was the default up to 0.0.33), upload, download, price_total, loc, id, or gpu |
--limit N | Limit the number of rows shown |
--format FORMAT | Output format: table (default) or json |
--json | Same as --format json (since 0.0.39) |
--country, --min-vram, --max-price, --tier, --min-cpus and the price_gpu default sort exist since Lium CLI 0.0.40 (--json since 0.0.39). 0.0.33 and earlier filter on --gpu, --count, --min-cuda and distance only.
Filters combine with AND. Since 0.0.40 the JSON output has top-level country_code and city for every node, so anything the flags cannot express can still be done with jq; in the same release disk_gb became the node's free disk (what a pod can get) and disk_total_gb the total, and the table shows CPUs and Disk free (Gb).
Examples
lium ls # List all nodes
lium ls --gpu H100 # Only H100 GPUs
lium ls --gpu H100 --count 8 --country US,NL --max-price 2.50 # since 0.0.40
lium ls --min-vram 80 --min-cuda 12.8 --tier secure # since 0.0.40
lium ls --min-cuda 12.4 # Only nodes supporting CUDA 12.4 or higher
lium ls --min-cpus 32 # At least 32 CPU threads (for a CPU-bound preprocessing stage) — since 0.0.40
lium ls --sort price_gpu # Cheapest $/GPU·h first
lium ls --limit 10 # Show only the first 10 rows
lium ls --format json # JSON output
Pareto-optimal choices in the table are marked with ★: a node keeps the ★ while no other listed node dominates it. Download speed is compared first: a node more than 10 % faster to download dominates outright, and one more than 10 % slower never does. When download speeds are within 10 % and the prices differ, the cheaper node per GPU-hour dominates when it is no worse (0.1 % tolerance) on every other column the CLI compares: VRAM, RAM, disk, PCIe, the GPU's reported graphics and memory clocks, upload, combined up+down bandwidth and US location. When the prices are within $0.01, a node dominates when it is more than 10 % better on combined bandwidth or upload, or is in the US while the other is not, and not more than 10 % behind on those three (the other columns are not looked at then); when those three tie, it dominates when it is better on any of VRAM, RAM, disk, PCIe or the GPU clocks and no worse on the rest. Nodes under 100 Mbps download are never ★. It is a hint, not a ranking; --sort replaces the default order, and the one-line legend under the table (since 0.0.40) says the same.
Tier column
Each node shows its Tier, so you can see the interruption risk before renting:
- secure — the default. If the rental is interrupted, the provider is penalized (their rental fee is withheld), so these nodes are the stable choice.
- spot — the provider can reclaim the node at any time with no penalty, and you are not compensated if the pod is interrupted. Use spot for cheaper, interruptible, fault-tolerant workloads.
With --format json, the same value is exposed as the tier field on each node ("spot", "secure", or null when unknown).
See Node Tier: Secure vs Spot for the full comparison.