GPU CLOUD / GHOSTVPS

GPU Cloud. Ideas in motion.

Train models, create with AI and run GPU workloads. Explore single and multi-GPU configurations, ready-made templates and usage-based billing.

PyTorchCUDAComfyUIJupyter
Explore your GPU setup
16GPU models
34configurations
1–8GPUs per instance
PyTorch × CUDA × ComfyUI

GPU / PAY AS YOU GO

Compare GPU configurations

Explore the published configurations and USD hourly rates. Choose a family or search the specifications to narrow your options.

34 matching configurations

These rows compare configurations. Opening Console does not preselect this row; choose the instance, region and operating system there, then review the final price before confirming.

Container disk: 30–200 GB. Persistent volume: 0 GB or 10–1,000 GB. Retained storage can continue billing while stopped.

Container disk is temporary and its files are lost on restart. The persistent volume at /workspace is retained across restarts.

Persistent volume: running 0.000232, stopped 0.000427 USD/GB/hour.

SOFTWARE CHOICES

Official GPU templates

These are current catalog templates, not a promise that every template works with every GPU or region. Console checks CUDA compatibility and capacity after you choose the GPU and location.

  • Pytorch 2.1.0
  • Pytorch 2.2.0
  • Pytorch 2.4.0
  • Pytorch 2.8.0
  • ComfyUI - CUDA 12.8
  • Pytorch 2.8.0 for clusters
  • Pytorch 2.9.0 for clusters
  • Network storage file browser
  • ComfyUI - CUDA 13.0
Compare GPU configurations
GPU modelGPU countTotal VRAMGPU compute USD / hourConfigure in Console
B300 MIG 34GB
1 listed locations
  • Washington, USA (US-WA-2)
1× GPU34 GB$0.6500Configure in Console
RTX 3090
1 listed locations
  • Czech Republic (EU-CZ-1)
1× GPU24 GB$0.6500Configure in Console
RTX PRO 4000
1 listed locations
  • Romania (EU-RO-1)
1× GPU24 GB$0.7410Configure in Console
A40
2 listed locations
  • Montreal, Canada (CA-MTL-1)
  • Sweden (EU-SE-1)
1× GPU48 GB$0.7670Configure in Console
L4
2 listed locations
  • Romania (EU-RO-1)
  • Iceland (EUR-IS-1)
1× GPU24 GB$0.8970Configure in Console
PRO 6000 MIG 24GB
3 listed locations
  • Iceland (EUR-IS-2)
  • Pennsylvania, USA (US-PA-1)
  • Washington, USA (US-WA-1)
1× GPU24 GB$0.8970Configure in Console
RTX PRO 4500
2 listed locations
  • Czech Republic (EU-CZ-1)
  • Romania (EU-RO-1)
1× GPU32 GB$0.9360Configure in Console
RTX PRO 4500 SE
1 listed locations
  • Kansas, USA (US-KS-2)
1× GPU32 GB$0.9360Configure in Console
RTX 4090
5 listed locations
  • Czech Republic (EU-CZ-1)
  • Romania (EU-RO-1)
  • Iceland (EUR-IS-1)
  • Iceland (EUR-IS-2)
  • Illinois, USA (US-IL-1)
1× GPU24 GB$1.1570Configure in Console
B300 MIG 34GB
1 listed locations
  • Washington, USA (US-WA-2)
2× GPU68 GB$1.3000Configure in Console
A40
1 listed locations
  • Montreal, Canada (CA-MTL-1)
2× GPU96 GB$1.5340Configure in Console
PRO 6000 MIG 24GB
3 listed locations
  • Iceland (EUR-IS-2)
  • Pennsylvania, USA (US-PA-1)
  • Washington, USA (US-WA-1)
2× GPU48 GB$1.7940Configure in Console
RTX 5090
1 listed locations
  • Iceland (EUR-IS-1)
1× GPU32 GB$1.8070Configure in Console
RTX PRO 4500
2 listed locations
  • Czech Republic (EU-CZ-1)
  • Romania (EU-RO-1)
2× GPU64 GB$1.8720Configure in Console
RTX PRO 4500 SE
1 listed locations
  • Kansas, USA (US-KS-2)
2× GPU64 GB$1.8720Configure in Console
RTX 4090
3 listed locations
  • Czech Republic (EU-CZ-1)
  • Romania (EU-RO-1)
  • Illinois, USA (US-IL-1)
2× GPU48 GB$2.3140Configure in Console
B300 MIG 34GB
1 listed locations
  • Washington, USA (US-WA-2)
4× GPU136 GB$2.6000Configure in Console
A100 SXM
3 listed locations
  • Iceland (EUR-IS-1)
  • Kansas, USA (US-KS-2)
  • MD, USA (US-MD-1)
1× GPU80 GB$2.8470Configure in Console
RTX PRO 6000
7 listed locations
  • Czech Republic (EU-CZ-1)
  • Iceland (EUR-IS-1)
  • Iceland (EUR-IS-2)
  • Colorado, USA (US-CO-1)
  • MD, USA (US-MD-1)
  • North Carolina, USA (US-NC-2)
  • Pennsylvania, USA (US-PA-1)
1× GPU96 GB$3.4970Configure in Console
PRO 6000 MIG 24GB
3 listed locations
  • Iceland (EUR-IS-2)
  • Pennsylvania, USA (US-PA-1)
  • Washington, USA (US-WA-1)
4× GPU96 GB$3.5880Configure in Console
RTX 4090
1 listed locations
  • Czech Republic (EU-CZ-1)
4× GPU96 GB$4.6280Configure in Console
B300 MIG 34GB
1 listed locations
  • Washington, USA (US-WA-2)
8× GPU272 GB$5.2000Configure in Console
A100 SXM
1 listed locations
  • MD, USA (US-MD-1)
2× GPU160 GB$5.6940Configure in Console
H100 SXM
9 listed locations
  • India (AP-IN-1)
  • Japan (AP-JP-1)
  • Montreal, Canada (CA-MTL-1)
  • France (EU-FR-1)
  • Iceland (EUR-IS-3)
  • Norway (EUR-NO-2)
  • Georgia, USA (US-GA-2)
  • Missouri, USA (US-MO-1)
  • Nebraska, USA (US-NE-1)
1× GPU80 GB$5.7070Configure in Console
H200 SXM
7 listed locations
  • Japan (AP-JP-1)
  • Iceland (EUR-IS-4)
  • Iceland (EUR-IS-5)
  • California, USA (US-CA-2)
  • Colorado, USA (US-CO-1)
  • Georgia, USA (US-GA-2)
  • North Carolina, USA (US-NC-1)
1× GPU141 GB$6.8770Configure in Console
RTX PRO 6000
3 listed locations
  • Iceland (EUR-IS-2)
  • North Carolina, USA (US-NC-2)
  • Pennsylvania, USA (US-PA-1)
2× GPU192 GB$6.9940Configure in Console
PRO 6000 MIG 24GB
3 listed locations
  • Iceland (EUR-IS-2)
  • Pennsylvania, USA (US-PA-1)
  • Washington, USA (US-WA-1)
8× GPU192 GB$7.1760Configure in Console
B200
2 listed locations
  • North Carolina, USA (US-NC-2)
  • Nebraska, USA (US-NE-1)
1× GPU180 GB$10.3870Configure in Console
A100 SXM
1 listed locations
  • MD, USA (US-MD-1)
4× GPU320 GB$11.3880Configure in Console
H100 SXM
2 listed locations
  • Georgia, USA (US-GA-2)
  • Nebraska, USA (US-NE-1)
2× GPU160 GB$11.4140Configure in Console
B300
1 listed locations
  • Iceland (EUR-IS-1)
1× GPU288 GB$11.6870Configure in Console
H200 SXM
3 listed locations
  • Iceland (EUR-IS-4)
  • Colorado, USA (US-CO-1)
  • Georgia, USA (US-GA-2)
2× GPU282 GB$13.7540Configure in Console
RTX PRO 6000
1 listed locations
  • North Carolina, USA (US-NC-2)
4× GPU384 GB$13.9880Configure in Console
A100 SXM
1 listed locations
  • Missouri, USA (US-MO-1)
8× GPU640 GB$22.7760Configure in Console

Rates checked in Console: 2026-11-19 23:31 UTC+03:00. This is a configuration snapshot; confirm the selected region, storage and final quote in Console.

Service-level conditions

These conditions describe the Console service entry, not every hourly hardware configuration. Verify the selected instance in Console.

Published plan conditions — GPU VPS
CPU allocationListed; review details
View published evidence
  • Live Usage Monitoring (GPU, VRAM, CPU, RAM and disk stats)
Traffic allowanceCheck in Console
Network portCheck in Console
SMTP port 25Check in Console
PTR / reverse DNSCheck in Console
BackupCheck in Console
System optionsCheck in Console
Software / Windows licence costCheck in Console
Configuration choicesCheck in ConsoleListed configuration choices do not establish an upgrade path after purchase.
Upgrades after purchaseCheck in Console
Service managementCheck in ConsoleA control panel does not by itself mean a managed service.

GHOSTVPS / GPU COMPUTE

Start with your workload. Choose your GPU.

For model training, inference, image generation and GPU processing, compare memory, GPU count and storage before deploying.

B300 MIG 34GBRTX 3090RTX PRO 4000A40L4PRO 6000 MIG 24GBRTX PRO 4500RTX PRO 4500 SE

These models are advertised in Console; availability, memory variants and configurations depend on current capacity.

VRAM for your model

Match GPU memory to your model, precision, batch size and working data. Check the memory of the exact GPU variant, not only its model name.

GPU count and software support

Choose a single or multi-GPU configuration offered in Console. Your software must support the selected GPU setup; more GPUs do not automatically improve every workload.

Space for models and results

Allow storage for model weights, datasets, templates and output. Review storage type, persistence and cost before deployment.

AI / DEVELOPMENT

Your tools, ready for the next experiment.

Choose an available AI template for your workflow. Framework and CUDA compatibility depend on the selected image and GPU.

Compare plans

Jupyter Notebook

Use the password-protected web notebook environment offered by the service to develop and run experiments.

SSH and network access

The service lists root SSH access and a dedicated public IPv4 for direct TCP access. Use the connection details and exposed ports shown in Console.

Monitoring and management

Review GPU, VRAM, CPU, RAM and disk usage, logs and cost reports. Template reinstallation is available; back up required data before reinstalling.

UNDERSTAND THE BILLING

Understand running and stopped costs.

The order amount is initial cloud credit, not the price of a GPU per hour. Review the selected GPU rate and billing interval in Console.

While running

GPU compute is charged for the whole time the instance is running, including idle time. Billing is not based only on moments when a model is processing. Storage and applicable extras can add to the total.

While stopped

Stopping the instance releases GPU compute, so GPU compute charges stop. A retained persistent volume remains billable at the applicable stopped rate; check the full stopped quote in Console.

Review storage retention before stopping, reinstalling or deleting. Keep an independent copy of important models and results; do not assume every operation preserves data.

GPU / CONSOLE

From an idea to a GPU workspace.

Add cloud credit, choose an available GPU and template, review the running and stopped rates, then create and manage the service in Console.

Configure in Console

Before you order

Know what fits before you choose.

Choose GPU memory and compute resources for your model. A stopped server can still incur storage or other charges; check the selected service’s running and stopped rates.

Need help choosing?
Is the initial cloud credit a GPU hourly price?

No. It is initial cloud credit. The GPU model, count, storage and billing interval determine the actual charge shown in Console.

What is charged when I stop the server?

The published stopped quote releases GPU compute. A retained persistent volume continues to be billed, and its stopped per-GB rate can differ from the running rate. Check the current configuration and quote in Console before stopping.

Are all listed GPUs always available?

No. GPU model, quantity, memory variant and location depend on available capacity. Confirm the exact configuration in Console before deployment.

Will my files survive a reinstall or deletion?

Do not assume so. Persistence depends on the selected storage and operation. Back up required files before reinstalling or deleting the service.

Can I keep my model on disk and pay for GPU only while it processes?

You can keep model files on a persistent volume, typically mounted at /workspace. GPU compute is billed while the instance is running, even when idle. Stop the instance to stop GPU compute charges; retained persistent storage continues to be billed. Container disk is temporary and should not be relied on to preserve model files. Check both running and stopped quotes in Console.