Massed Compute GPUs on DeployGPU

The Massed Compute offers on DeployGPU are GPU virtual machines in the United States from Massed Compute, launched on DeployGPU's own Massed Compute account and billed from your DeployGPU balance. You do not need a Massed Compute account. The one detail that trips people up is the login: the SSH user is Ubuntu, with a capital U.

Offers now

48

From

$0.55/GPU-hr

GPUs per VM

1x, 2x, 4x, 8x

GPU models

8

United States · checked 2026-10-13 13:51 UTC

Massed Compute GPUs you can launch now

Each row is one GPU family in stock on Massed Compute right now, with the number of VM offers, the GPU counts per VM and the cheapest price per GPU-hour, refreshed every 15 minutes. The prices are what you pay on DeployGPU, including any platform fee.

For Massed Compute's prices next to the rest of the market, see Massed Compute on GpuPerHour.

How a Massed Compute VM differs from the other 3 VM providers

Log in as Ubuntu with a capital U on port 22

Linux usernames are case-sensitive, and Massed Compute's VMs use Ubuntu, not ubuntu. Typing the lowercase name from habit, or from a guide written for another cloud, fails with Permission denied (publickey) even when your key is right. Copy the command from the deployment page:

ssh Ubuntu@YOUR_VM_IP -p 22

The same capital U goes into scp, rsync, VS Code's Remote-SSH config and any SSH tunnel:

rsync -avP Ubuntu@YOUR_VM_IP:~/results/ ./results/
ssh -N -L 8888:127.0.0.1:8888 Ubuntu@YOUR_VM_IP -p 22

If the login still fails with the right user, Permission denied (publickey) covers the key side, and the SSH guide covers keys from the start.

A fixed disk and no volumes

The disk on a Massed Compute VM is fixed per offer. There is no size to pick at launch and no volume to attach, so the disk in the offer is all the space you get for model weights, Docker images, datasets and outputs. Check the Disk column before you launch a model-heavy job, and see no space left on device if it fills up.

Stopping deletes the VM and its disk, and with no volumes on Massed Compute, nothing survives a stop unless you copied it off first. The stop guide has the checklist, and moving data to and from a cloud GPU covers rsync, rclone and the Hugging Face Hub. If you need storage that outlives the VM, Verda and Hyperstack offer volumes.

Templates, the API and what you pay

All four DeployGPU templates run on Massed Compute: JupyterLab with PyTorch, ComfyUI, Open WebUI with Ollama and the plain GPU VM. Template apps open on a private HTTPS link that only your DeployGPU login can use. Through the API you launch plain VMs on Massed Compute with the same calls as on the other providers.

The hourly price is the rate you are billed, and launching reserves the first hour. When you stop, the unused seconds of the current hour go back to your balance, and a launch that fails is refunded in full. Card top-ups carry a 5% + $0.30 processing fee. The cost calculator works out a run from live prices.