11 GPUs available now

GPU cloud for AI.
Instances. Serverless. Clusters.

GPU instances for training. Serverless for inference. Clusters for scale. 44 models. Per-second billing. From $0.18/hr.

Deploy an Instance Go Serverless GPU Clusters

Three ways to deploy

Single GPU instances. Auto-scaling serverless. Multi-node clusters. All per-second billing.

GPU Instances

  • Persistent GPU instance
  • Deploy in ~30 seconds
  • Ollama API, Docker, or SSH
  • Pause and resume anytime
  • Spot pricing — 50% off

Best for: training, dev, notebooks

Browse GPUs

Serverless

  • Auto-scaling inference endpoint
  • Scale to zero — $0 when idle
  • 970+ pre-built templates
  • ~2 second cold start
  • Sync + async execution

Best for: inference APIs, chatbots, batch

Go Serverless

Clusters

  • Multi-node GPU clusters
  • 2–32 instances with InfiniBand
  • Up to 8 GPUs per node
  • Distributed training at scale
  • Network storage included

Best for: large models, distributed training

Deploy Cluster

USE CASES

Built for every AI workload

Inference

Deploy models as auto-scaling APIs. Scale to zero. $0 when idle. Sub-second cold starts.

Learn more →
Training & Fine-Tuning

Single GPU to 256-GPU clusters. Per-second billing. PyTorch, Axolotl, DeepSpeed.

Learn more →
AI Agents

Zero-friction provisioning. API key in seconds. Pay per second. Built for agent-first workflows.

Learn more →

As simple as git

computegpu deploy is like git push. Three commands to go from zero to running GPU.

npm i -g computegpu
computegpu signup you@co.com my-agent
computegpu deploy --gpu "RTX 4090"
1

Install

npm i -g computegpu — one command, zero config.

2

Sign up

Get a gpu_ API key instantly. Verify with $1 (credited to wallet).

3

Deploy

GPU running in ~30 seconds. Per-second billing. Stop anytime.

RTX 4090 RTX 5090 A100 H100 H200 B200 All GPUs

GPU pricing

Per-second billing. Deploy instantly. Stop anytime.

High Performance
H100 SXM
80 GB HBM3
Spot:
Deploy
H200
141 GB HBM3e
Spot:
Deploy
A100 80GB
80 GB HBM2e
Spot:
Deploy
RTX 4090
24 GB GDDR6X
Spot:
Deploy
Serverless (scale to zero)
Ollama
Any model, any GPU
from $0.39/hr
$0 when idle
Deploy
vLLM
OpenAI-compatible
from $0.85/hr
$0 when idle
Deploy
SDXL
Image generation
from $0.39/hr
$0 when idle
Deploy
Whisper
Audio transcription
from $0.18/hr
$0 when idle
Deploy
All 44 GPU instances → Serverless API docs →
Per-second billing

No hourly rounding. Instances bill per second of uptime. Serverless bills per second of compute. Stop anytime.

Scale to zero

Serverless endpoints cost $0 when idle. Auto-scale under load, scale down when quiet. No manual start/stop.

CLI-first

As simple as git. npm i -g computegpu then computegpu deploy --gpu "RTX 4090". Full REST API included.

Available GPUs

44 models · 11 available now
NVIDIA GeForce
RTX 3060
12 GB
0 available
RTX 3070
8 GB
0 available
RTX 3080
10 GB
0 available
RTX 3090
24 GB
1 available
RTX 4060
8 GB
Coming soon
RTX 4060 Ti
16 GB
0 available
RTX 4070
12 GB
Coming soon
RTX 4070 Ti
12 GB
0 available
RTX 4080
16 GB
0 available
RTX 4090
24 GB
1 available
RTX 5070
12 GB
1 available
RTX 5080
16 GB
1 available
RTX 5090
32 GB
1 available
Datacenter
A40
48 GB
0 available
L4
24 GB
1 available
L40S
48 GB
1 available
A100 40GB
40 GB
Coming soon
A100 80GB
80 GB
0 available
H100
80 GB
1 available
H100 NVL
94 GB
1 available
H200
141 GB
1 available
B100
180 GB
Coming soon
B200
180 GB
1 available
B300
288 GB
0 available
GB200
384 GB
Coming soon
GB300
576 GB
Coming soon
DGX H100
640 GB
Coming soon
DGX B200
1440 GB
Coming soon

Developer tools

Deploy instances and serverless endpoints from your terminal or code.

CLI
npm i -g computegpu
computegpu deploy --gpu RTX4090
computegpu stop POD_ID
CLI docs →
Instance API

Deploy, pause, resume, destroy GPU instances. Full lifecycle via REST.

Instance API docs →
Serverless API

Create endpoints, run sync/async inference, poll results. Scale-to-zero.

Serverless docs →
Templates

PyTorch, vLLM, ComfyUI, Whisper, Jupyter. Pick and launch.

Browse templates →

Three commands to GPU

npm i -g computegpu && computegpu deploy --gpu "RTX 4090"

Get started free View docs
Help
Welcome to ComputeGpu

Rent cloud GPUs instantly. 44 models from $0.20/hr. Per-second billing. No commitments.

Getting Started
  1. Sign up — Create an account (free, no credit card)
  2. Add credits — Top up your wallet ($5 minimum via Stripe)
  3. Deploy a GPU — Browse GPUs, click Rent, get your endpoint in ~30 seconds
What You Get
Why ComputeGpu?
Developer Tools

CLI: npm i -g computegpu

API: Full reference | GPU Cloud API