You see the provider's price. We add 10%. That's it. No bandwidth fees, no egress charges, no minimum spend. You buy credits with a card, credits get charged hourly while your instance runs.
01 / deploy
From I need an H100 to ssh root@… in 60 seconds.
01 Browsemarketplace
H100
$1.90/hr
80GB VRAM
available now
02 Configuretemplate
⚡PyTorch
🦙Llama
🖼ComfyUI
📓Jupyter
⬛Ubuntu
⚙Custom
Deploy →
03 Connectterminal
$ ssh [email protected]
Welcome to DeployGPU (Ubuntu 22.04)
CUDA 12.4 · PyTorch 2.3 · 24GB VRAM
>>> import torch
>>> torch.cuda.is_available()
True
>>>
no SSH config /no CUDA roulette /no three invoices /no lock-in
02 / included
Everything between you and your model — handled.
Managed SSH
One key. Every instance. Every provider.
PyTorch
Llama.cpp
ComfyUI
Jupyter
Preconfigured CUDA
Drivers matched. No Dockerfile.
vast.aiH100-$1.47/hr
verdaV100-$0.14/hr
hyperstackL40S-$0.89/hr
balance$47.20
Unified Billing
One balance. One invoice.
vast.ai
verda
hyperstack
latitude
massed compute
Provider Failover
One unavailable? Deploy on another.
$ scp -r root@gpu:~/weights ./local/
> transferred 4.2 GB ✓
Zero Lock-in
Your data. Export or leave any time.
five things handled. one thing left: your model.
03 / marketplace
Every GPU. Every cloud. Live prices.
Updated every 15 minutes from the actual market.
| GPU Family | VRAM | From▲ | Available |
|---|---|---|---|
| A6000 | 48 GB | from$0.49/hr | |
| RTX 4090 | 24 GB | from$0.67/hr | |
| RTX 6000 Ada | 48 GB | from$0.79/hr | |
| L40 | 48 GB | from$0.84/hr | |
| L40S | 48 GB | from$0.88/hr | |
| A100 | 80 GB | from$1.00/hr | |
| RTX 5090 | 32 GB | from$1.07/hr | |
| H100 | 94 GB | from$1.90/hr | |
| H200 | 141 GB | from$4.85/hr | |
| B200 | 192 GB | from$4.89/hr | |
| B300 | 262 GB | from$6.99/hr | |
| GB300 | 288 GB | — | |
| RTX PRO 6000 | 96 GB | — | |
| V100 | 32 GB | — |
A6000$0.49/hr48 GB
RTX 4090$0.67/hr24 GB
RTX 6000 Ada$0.79/hr48 GB
L40$0.84/hr48 GB
L40S$0.88/hr48 GB
A100$1.00/hr80 GB
RTX 5090$1.07/hr32 GB
H100$1.90/hr94 GB
H200$4.85/hr141 GB
B200$4.89/hr192 GB
B300$6.99/hr262 GB
GB300—288 GB
RTX PRO 6000—96 GB
V100—32 GB
05 / reserve
Reserve a GPU Cluster
Submit your requirements and we'll get you quotes from multiple providers within 24 hours. From single-node to multi-node clusters — we handle the sourcing.
Quotes from multiple providers within 24 hours
NVLink & InfiniBand cluster configurations
Flexible terms — days to 12 months
Dedicated support for cluster deployment
Volume pricing on reserved capacity
04 / faq
Common questions.
If a provider terminates your instance (hardware failure, maintenance), we detect it within 5 minutes and mark it interrupted. Unused credit for that hour is refunded automatically. You redeploy on any provider in one click.
Your instance runs on the provider's hardware. You hold the private SSH key — once provisioned, we have no way to access your machine. When you terminate, the instance is destroyed and we don't retain your data.
Yes. API keys (Bearer gpu_live_*) give you programmatic access to deploy, terminate, and list instances. Same capabilities as the dashboard.
Vast.ai, Verda, Hyperstack, Latitude, and Massed Compute today. More on the way. You deploy to any of them from the same dashboard, same billing, same SSH flow.
None. No credit card to sign up. Buy credits when you're ready. Minimum top-up is $5. Instances bill by the hour, terminate any time.
Ready to deploy?
Pick a GPU, choose a template, and be running in under 60 seconds.
Start DeployingNo credit card required
Ready to deploy?
Pick a GPU, choose a template, and be running in under 60 seconds.
Start DeployingNo credit card required
06 / enterprise
Need GPUs managed end-to-end?
We handle the infrastructure so your team can focus on the model. From provisioning to monitoring — one point of contact.
Dedicated account manager
Custom templates & infrastructure
Volume pricing & reserved capacity
SLA-backed uptime
Priority support (Slack / Discord)