Rent an RTX 5090 with Python
The Lium Python SDK rents GPUs from code: pip install lium.io, set LIUM_API_KEY, and a few lines find an RTX 5090, start a pod, run your command and remove the pod.
No RTX 5090 is free right now. Reference price $0.40 per GPU-hour. Live RTX 5090 prices
pip install lium.io, then
from lium.sdk import Lium
lium = Lium() # reads LIUM_API_KEY
node = min(lium.ls(gpu_type="RTX5090"), key=lambda n: n.price_per_hour)
pod = lium.up(executor_id=node.id, name="rtx5090-job")
try:
ready = lium.wait_ready(pod, timeout=600)
if ready is None:
raise RuntimeError("pod not ready after 10 minutes")
print(lium.exec(ready, command="nvidia-smi")["stdout"])
finally:
for p in lium.ps():
if p.id == pod["id"]:
lium.rm(p)Or run one function on the GPU
The @lium.machine decorator rents the cheapest matching node, ships a Python function to it, installs its requirements, runs it, returns the result and removes the pod. You call the function like any other.
Which images work on an RTX 5090
The RTX 5090 is a Blackwell card and needs CUDA 12.8 or newer. Pick a template built on CUDA 12.8 or later, or filter nodes with lium ls --min-cuda 12.8.
Frequently asked questions
Where are the SDK docs?
docs.lium.io/developers/sdk covers the client and the decorator, with runnable examples for the pod lifecycle, training, custom templates and a vLLM server.
How is a pod billed?
Every pod has an hourly price and is billed per second, with no minimum rental period. Billing stops when the pod is removed. A pod rented with --ttl removes itself when the time is up.
Can an AI agent do this without a browser?
Yes. Signup, API keys, renting, running commands and removal all work from the CLI or the REST API with no browser and no e-mail confirmation. docs.lium.io/developers/agents is the one-page guide.
For AI agents — CLI install, agent skill, headless signup: llms-full.txt