GPU rental + model APIs · one balance

CloudGPU: GPU rental and model APIs on one balance.

CloudGPU rents on-demand GPUs such as the RTX 5090, RTX 4090 and A100 by the minute, and serves popular models through an OpenAI-compatible API, both paid from one prepaid balance by card or USDT.

Claude models on CloudGPU: see our per-1M-token rates next to the official list prices (checked 2026-10-24) on Pricing. CloudGPU is an independent service, not the official Claude API.

50 models across 5 modalities · per-token model APIs · per-minute GPU billing

check_circleUSDT TRC-20 top-up
check_circleOpenAI-compatible
check_circleRoot SSH and JupyterLab
check_circleMinute or hourly billing

Compute infrastructure for AI agents

Start with your project and expected result. Your agent checks task fit, shows a quote and asks you to approve the budget before running.

  1. 1

    Agent checks task fit and live GPU availability.

  2. 2

    You sign in with Google, add funds if needed and approve limits.

  3. 3

    Agent runs the approved task, saves results and verifies release.

Read the agent guide

Requires an agent with HTTPS tools or a configured MCP client. Connection approval is not unlimited spending permission.

Choose a GPU workflow

Read the setup guide, then choose a template and check current options on the deployment page.

ComfyUI / FLUX image workflows

Prepare workflows, model permissions and result downloads before choosing a GPU.

MiniMax H3 video workflows

Review the setup instructions and examples before selecting the video template.

Open-source model inference

Prepare your model, review resources and billing, then connect your inference service.

LoRA and QLoRA fine-tuning

Start with data and memory requirements, configure training and export adapters.

Two ways to use the GPUs

Same cards, same price. You pick how much of the machine you want to think about.

For AI Agent Developers

Per Token or Per Minute.
Deploy Your Own AI Models.

Run DeepSeek, Qwen, Llama and Mistral on dedicated GPUs. You pay for GPU time rather than tokens. The billing period and minimum charge are shown before you confirm a rental. Review the quote for your selected configuration.

How it works

smart_toy

1. Pick a Model

Choose from DeepSeek, Qwen, Llama, Mistral and more. Or rent a raw GPU for full control.

touch_app

2. Click Deploy

We handle GPU allocation, environment setup, model download — everything.

api

3. Get Your API

Paste the endpoint into your code. No per-token charges; you pay for GPU time. Works with any OpenAI-compatible SDK.

GPU prices

Transparent pricing. No hidden fees. No egress costs.

GPU ModelCloudGPU
RTX 2080$0.15/hr
RTX 2080 Ti$0.16/hr
RTX 3080$0.19/hr
RTX 3080 Ti$0.20/hr
A30$0.22/hr
V100 16G$0.25/hr
RTX 3090$0.29/hr
RTX 4080 SUPER 32G$0.33/hr
V100 32G$0.35/hr
RTX 4090D$0.37/hr
RTX 4090$0.39/hr
A40$0.39/hr
RTX A6000$0.49/hr
RTX 4090D 48G$0.59/hr
RTX 5090D$0.65/hr
RTX 4090 48G$0.69/hr
RTX 5090$0.69/hr
A800 40G$0.69/hr
A100 40G$0.79/hr
Ascend 910B$0.95/hr
RTX PRO 5000 72G$1.04/hr
A800 80G$1.19/hr
NVIDIA RTX PRO 5000 72G$1.21/hr
H20 96G$1.29/hr
RTX 6000D$1.29/hr
RTX PRO 6000 Max-Q$1.39/hr
A100 80G$1.59/hr
H800$1.79/hr
RTX PRO 6000 96G$1.79/hr
H20-3e 141G$1.89/hr

CloudGPU is in public beta. There is no SLA: no uptime guarantee and no service credits. The balance does not expire. The billing period and minimum charge are shown before you confirm a rental. Review the quote for your selected configuration.

* No hidden fees. No egress charges. Stop anytime.

Ready to start?

Pay by Alipay, Google Pay, Apple Pay, card or USDT.

Pick a template and deploy your first model.

Frequently asked questions

What is CloudGPU?

CloudGPU (cloudgpu.app) rents on-demand GPUs such as the RTX 5090, RTX 4090 and A100 by the minute, and serves popular models through an OpenAI-compatible API. Both are paid from one prepaid balance.

How is GPU time billed?

Per-minute options have a 1-minute minimum; unused prepaid minutes are refunded when you destroy the instance. Some configurations use whole-hour billing with a 1-hour minimum; unused time within a charged hour is not refunded on voluntary stop. The billing period, minimum charge and initial prepayment are shown before confirmation. No subscription.

Is the API OpenAI-compatible, and what is the base URL?

Yes. Set the base URL to https://cloudgpu.app/v1 and use your CloudGPU API key in the OpenAI SDK or any tool that accepts an OpenAI-compatible endpoint.

How can I pay?

Top up one prepaid balance by card or with USDT on TRC-20. Businesses can also pay by bank transfer. Methods and minimums are listed at https://cloudgpu.app/payments.

What happens to my data when I destroy an instance?

Destroying an instance, including when its balance runs out, erases its local disk. Eligible funded accounts can explicitly save project files in Workspace and restore a saved version in another Job or apply it to a running GPU. Saved files remain subject to storage quotas and retention rules. Workspace does not automatically back up the disk or collect results; save or download results before destroying the instance.