AmpleRun

GPU marketplace

Deploy any workload in one click. Or earn with your machine.

NVIDIA CUDA, AMD, and Apple silicon.

Deploy a modelStart hosting

No account needed to browse. Pay only while it runs.

From$0.06/hlive, per GPU
GPUs available
Countries4
Median price
GPU-hours served

Open models. One click.

Ready-made images from the catalog. Pick one and deploy it.

Names and logos belong to their owners. A mark here is not a partnership.

How to deploy

  1. 1Pick a model84 templates
  2. 2Press DeployOne click. Ready in about 11 s.1
  3. 3It runsBilled by the millisecond once it's ready.

GPUs listed now

Price per GPU-hour, rounded up to the cent. The bar compares that price with the dearest row here.

GPUPriceListed
Apple M1 Pro (14-core GPU)Apple silicon$0.06/h1
Apple M4 Pro (20-core GPU)Apple silicon$0.12/h1
Apple M3 Max (30-core GPU)Apple silicon$0.13/h1
Apple M5 Max (40-core GPU)Apple silicon$0.26/h1

All GPUs

Pick a model. Press Deploy.

We match the best GPU for the price.

Qwen3 1.7B

Published

Runs on 1× Apple M1 Pro (14-core GPU) · New host · about $0.06/h

Runs 1 h, then stops

Change

Any GPU, any budget.

Chat, video, or your own container. NVIDIA CUDA, AMD, and Apple silicon.

From$0.06/hup to$0.26/hPer GPU. 4 listed right now.

  • NVIDIA CUDA
  • AMD
  • Apple silicon
  • Intelbeta
People runLLMs and agents 28Notebooks and ML 9OS images and tools 7Image generation 4Speech and transcription 2
Listed nowApple M5 Max (40-core GPU)Apple M3 Max (30-core GPU)Apple M4 Pro (20-core GPU)Apple M1 Pro (14-core GPU)

Terms at a glance.

You pay the price on Deploy, only while it runs.

Billing
By the millisecond once it's ready. A failed start costs nothing.
Other sites
Busy on other marketplaces? We pause your listing here.
Readings
Load, memory, power and heat. A reading a GPU lacks says so.
Agents
Wallet sign-up, x402 payments, MCP and SDK. Agent docs · SKILL.md

How this compares

Hardware, billing, and how you start. Each page names its source and the day it was checked.

Questions

How do I run a model?
Pick one on this page and press Deploy. It starts on the cheapest GPU that fits. When it's running, press Chat now.
What does it cost?
Hosts set their own prices. The figure on Deploy is the hourly price. You pay only while it runs. See pricing.
Can my GPU earn money?

Yes. Paste one command on your machine. You earn from every rental.

Pick Dedicated, Idle or Co-listed. Idle and Co-listed rentals are Interruptible.

How do payouts work?

Withdraw unused credit from Billing. The minimum and the fee show before you confirm.

Disputed amounts wait until they're resolved.

What do I need to host?
  • Linux with NVIDIA CUDA or AMD, or a Mac with Apple silicon
  • Containers: NVIDIA, Docker and a public endpoint
  • AI models: no inbound port needed

The installer checks first and changes nothing if your machine doesn't qualify.

How is this different from Vast.ai?

You pick a template and deploy in one click. Vast's pricing docs describe their meter (pricing docs, checked 25 September 2026). Full comparison

Is there a minimum time?
No. It's billed by the millisecond once it's ready. Unused credit can be withdrawn.
Can I use my own Docker image?
Yes: any image pinned by digest, or a published template. It runs without root, with system files read-only and a writable job folder.

Deploy a workload, or earn with your machine.

Deploy a modelStart hosting
Sources and notes
  1. About 11 s from pressing Deploy to a model serving, measured on 7 Oct 2026. It was one measurement and the method isn't on file, so it varies by model and GPU.