ParalonCloud

GPUs available now, billed by the minute

Rent a GPU by the minute

Verified NVIDIA GPUs, from RTX 4090 and 5090 to A100 and H100, with JupyterLab and root access. No contract, stop any time, and spot machines at a discount.

Browse available GPUs See the price list
$0.03/h
From, right now
~1 min
To a running notebook
$0
Egress fees

Need 8 or more GPUs for a month or longer? See reserved GPUs

Live

GPU rental prices, right now

What each model costs per hour on the network this minute. Providers set their own prices, so the same card can be listed at more than one; the table shows the lowest.

See every machine
GPUVRAMAvailabilityPrice per hour
Rent RTX A6000Ampere48 GBAvailable now$0.37
Rent RTX 5090Blackwell32 GBAvailable now$0.75
Rent Quadro M6000Maxwell24 GBAvailable now$0.03
Rent RTX 5060 Ti16 GBAvailable now$0.18
Rent Tesla P100-PCIE-16GB16 GBAvailable now$0.10
Rent RTX 5070 Ti Laptop GPU12 GBAvailable now$0.16
56 more models, none available right now
Rent B300Blackwell Ultra288 GBNone right nowlist$9.40
Rent H200Hopper141 GBNone right nowlist$4.30
Rent H200 NVL140 GBNone right nowlist$4.20
Rent RTX PRO 6000 Blackwell Workstation Edition95 GBNone right nowlist$2.00
Rent RTX PRO 6000 Blackwell Server Edition95 GBNone right nowlist$2.00
Rent H100 SXMHopper80 GBNone right nowlist$2.80
Rent H100 PCIeHopper80 GBNone right nowlist$2.40
Rent A100 80GBAmpere80 GBNone right nowlist$2.40
Rent A100 80GB PCIeAmpere80 GBNone right nowlist$2.30
Rent H100 80GB HBM380 GBNone right nowlist$2.20
Rent L40SAda Lovelace48 GBNone right nowlist$1.10
Rent A40Ampere48 GBNone right nowlist$1.00
Rent L40S 48GB48 GBNone right nowlist$0.90
Rent A100 40GBAmpere40 GBNone right nowlist$1.40
Rent A100-SXM4-40GB40 GBNone right nowlist$1.40
Rent V100 32GBVolta32 GBNone right nowlist$0.12
Rent V100-PCIE-32GBVolta32 GBNone right nowlist$0.08
Rent RTX 5090 D32 GBNone right nowlist$0.60
Rent RTX A5000Ampere24 GBNone right nowlist$0.19
Rent RTX 4090Ada Lovelace24 GBNone right nowlist$0.48
Rent RTX 3090Ampere24 GBNone right nowlist$0.36
Rent RTX 3090 TiAmpere24 GBNone right nowlist$0.38
Rent RTX 5090 Laptop GPU24 GBNone right nowlist$0.38
Rent RTX A4500Ampere20 GBNone right nowlist$0.18
Rent V100-SXM2-16GBVolta16 GBNone right nowlist$0.12
Rent V100 16GBVolta16 GBNone right nowlist$0.09
Rent RTX A4000Ampere16 GBNone right nowlist$0.14
Rent RTX 5070 TiBlackwell16 GBNone right nowlist$0.27
Rent RTX 4080Ada Lovelace16 GBNone right nowlist$0.34
Rent RTX 4070 Ti SUPERAda Lovelace16 GBNone right nowlist$0.26
Rent Quadro P5000Pascal16 GBNone right nowlist$0.07
Rent RTX 4080 SUPER16 GBNone right nowlist$0.33
Rent RTX 508016 GBNone right nowlist$0.33
Rent RTX 4090 Laptop GPU16 GBNone right nowlist$0.30
Rent RTX 5070Blackwell12 GBNone right nowlist$0.25
Rent RTX 4070 TiAda Lovelace12 GBNone right nowlist$0.24
Rent RTX 3080 TiAmpere12 GBNone right nowlist$0.21
Rent RTX 3060Ampere12 GBNone right nowlist$0.08
Rent RTX 4070 SUPER12 GBNone right nowlist$0.19
Rent RTX 407012 GBNone right nowlist$0.17
Rent RTX 2080 TiTuring11 GBNone right nowlist$0.07
Rent RTX 3080Ampere10 GBNone right nowlist$0.15
Rent RTX 4060 TiAda Lovelace8 GBNone right nowlist$0.15
Rent RTX 4060Ada Lovelace8 GBNone right nowlist$0.14
Rent RTX 3070Ampere8 GBNone right nowlist$0.15
Rent RTX 2080Turing8 GBNone right nowlist$0.10
Rent RTX 3070 Ti8 GBNone right nowlist$0.17
Rent RTX 50608 GBNone right nowlist$0.16
Rent RTX 4070 Laptop GPU8 GBNone right nowlist$0.15
Rent RTX 30508 GBNone right nowlist$0.14
Rent RTX 4060 Laptop GPU8 GBNone right nowlist$0.13
Rent RTX A10008 GBNone right nowlist$0.12
Rent RTX 3060 Ti8 GBNone right nowlist$0.09
Rent Tesla P48 GBNone right nowlist$0.04
Rent RTX 3050 6GB Laptop GPU6 GBNone right nowlist$0.07
Rent RTX 3050 Ti Laptop GPU4 GBNone right nowlist$0.08

Benchmarked by ParalonCloud before listing

What you get

Know what you are renting before you rent it

The whole machine

Every GPU in the machine is yours for the rental, with its CPU, RAM and disk. Nothing is shared with another tenant.

JupyterLab in about a minute

A workspace opens in your browser with a terminal and root in the container. No drivers to install, no image to build.

Billed by the minute

You pay from prepaid credits for the minutes the machine runs. No contract, no minimum, stop any time.

Hardware we measured

Each GPU is benchmarked before it is listed and again after every reboot. The measured compute and memory bandwidth are on the machine's page.

A score for every machine

Machines and providers carry a reliability score from their finished rentals, with uptime and time to start.

No egress fees

Pull datasets and push checkpoints without a bandwidth bill on top of the GPU price.

How it works

From nothing to a running GPU in three steps

  1. 01

    Pick a GPU

    Browse the machines that are free right now: real specs, measured performance, location, the hourly price.

  2. 02

    Start it

    One click starts a JupyterLab workspace on the machine. It is usually ready in about a minute.

  3. 03

    Work, then stop

    Billing runs by the minute while it is up and ends when you stop it. Unused credits stay in your balance.

Two ways to rent

On-demand, or spot for less

On-demand

Yours until you stop it

  • The provider's listed price
  • Runs as long as you need
  • Counts towards the machine's reliability score

Spot

Cheaper, and reclaimable

  • Discounted from the list price
  • The owner may take it back with two minutes' notice
  • Billed only for the time you used

Use cases

What a rented GPU is for

01

Fine-tuning and training

Bring a dataset, train on one GPU or several in the same machine, and keep the checkpoint.

02

Running your own model

Serve a model you chose, with the runtime you chose: vLLM, llama.cpp, Ollama or your own code.

03

Image and video generation

Diffusion pipelines that need a full card's memory, without sharing it.

04

Research and notebooks

A GPU for an afternoon: open the notebook, run the experiment, stop the machine.

05

Batch jobs on spot

Work that can be resumed runs cheaper on spot machines, which the owner may take back with two minutes' notice.

06

From code and agents

Start and stop rentals over HTTP with the Rental API, or from Claude Code, Claude Desktop and Cursor through the MCP server.

Rental API docsMCP serverHow reliability is scored

FAQ

Renting a GPU, answered

How is a GPU rental billed?

By the minute, from prepaid credits, only while the machine is running. There is no contract and no minimum term: stop the rental and billing stops. Credits you have not used stay in your balance.

What do I get on the machine?

The whole machine for the length of the rental: every GPU in it, its CPU, RAM and disk. A JupyterLab workspace opens in your browser with a terminal and root in the container, so you can install what you need.

What is a spot GPU?

A spot GPU is a discounted machine whose owner may take it back at any time with two minutes' notice. You are billed only for the time you used. It suits work that can be checkpointed and resumed; on-demand machines are yours until you stop them.

How do I know the hardware is what it says?

Every GPU is benchmarked by ParalonCloud before it is listed and again after each reboot. The measured compute and memory bandwidth are shown on the machine's page next to what the card's specification says, and machines that pass carry a Verified mark.

How reliable are the machines?

The hardware belongs to independent providers, so each machine and each provider carries a reliability score built from its finished rentals, with uptime and the typical time to start. You can see it before you rent.

How do I pay?

Add credits by card or in USDC, then spend them on rentals and on the inference API. There is no subscription.

Can I rent a GPU from code?

Yes. The Rental API lists the machines available now, starts a rental, returns its connection details and stops it. The same actions are available to Claude Code, Claude Desktop and Cursor through the MCP server.

What happens to my data when the rental ends?

The rental's disk is cleared when it ends. Copy out checkpoints and results you want to keep before you stop the machine.

Your GPU is a minute away

Pick a machine, start it, and pay for the minutes it runs.

Browse available GPUs Add credits