Skip to content
Kilnara

Real GPU Compute
Without the Server Upkeep

Spin up an isolated GPU machine for a build, a model run or the idea you need to test today. SSH in like normal. Pay in ETH on Robinhood Chain, only while it runs.

Get Started See How It Works
Loading live pricesCA: Launch soon @Kilnaraxyz

Secure GPU Workspaces on Demand

Every workspace is an isolated machine started for one wallet. You connect with standard SSH keys, and access ends when the lease does. Nothing is left running and nothing keeps billing.

Intelligent Provisioning for Your Stack

SSHDockerCUDAPyTorchJupyterVS CodeSSHDockerCUDAPyTorchJupyterVS Code
+ Ephemeral Workspaces+ Real GPU Compute+ Pay Only While It Runs+ Ephemeral Workspaces+ Real GPU Compute+ Pay Only While It Runs+ Ephemeral Workspaces+ Real GPU Compute+ Pay Only While It Runs+ Ephemeral Workspaces+ Real GPU Compute+ Pay Only While It Runs

Spin up a developer machine for a build, a model run or the idea you need to test today. Your terminal, editor and containers stay the same. You pay in ETH on Robinhood Chain, and only while the machine runs.

Kilnara by the Numbers

GPUs Available Now

Latest Block

Hourly Price by Model

GPU Models

VRAM Tiers

Countries

Less Manual Work. More Intelligent Execution.

// 001

Smart Provisioning

Pick a workload template, or name the exact GPU, CPU, memory and region your experiment needs.

Browse Machines
// 002

Native Connectivity

Connect over plain SSH. Your terminal, editor, containers and the rest of your workflow stay exactly as they are.

Read the SSH Guide
// 003

Ephemeral Release

When the session ends, the machine is gone. No idle bill and no permanent server to babysit.

See the Lease Lifecycle

Start From a Workload Template

Pick a ready setup for the job, or bring any public container image and GitHub repo. The machine boots with it installed.

Bare CUDA

Root Over SSH on a Clean Box

PyTorch 2.5•CUDA 12.4•nvidia-smi
Launch Bare CUDA

JupyterLab

Notebooks in the Browser

Port 8888•PyTorch•Your repo cloned
Launch JupyterLab

ComfyUI

Node Graphs for Image Models

Stable Diffusion•Flux•Port 8188
Launch ComfyUI

vLLM

OpenAI-Compatible Serving

/v1/chat/completions•Port 8000•Any open model
Launch vLLM

Ollama

Open LLMs Behind an API

Llama 3.2 pulled•Port 11434•HTTP API
Launch Ollama

Fine-Tuning

LoRA and QLoRA Runs

Axolotl•PEFT•TRL
Launch Fine-Tuning

Blender

Headless Batch Renders

Cycles•OptiX•scp frames home
Launch Blender

Game Server

Dedicated Servers on SteamCMD

SteamCMD•Root access•Any title
Launch Game Server

Earn Kilnara Points

The launch event has ten challenges. Harder ones pay more and have fewer slots. Points are off-chain, tracked per wallet, and never move funds.

Get Started

Active Workspaces Every machine your wallet paid for, with live SSH details.
Model Marketplace Buy and sell models, datasets and templates for ETH.
Hardware Benchmarks Compare cards on inference, training and memory speed.
Platform Rewards Challenges and hackathon missions that pay in points.

Configure. Connect. Release.

Set the hardware, check the ETH quote and pay on Robinhood Chain. Connect with SSH or VS Code Remote. Export your output before the lease runs out.

The Right Weight for the Work

A10GL4
A10G class • none available right now

Inference and Mid-Tier Training

View Specs 24 GB class
40905090
RTX 4090 class • none available right now

Rapid Prototyping and General-Purpose Compute

View Specs 24 GB VRAM
H100H200
H100 class • none available right now

Large-Scale Modeling and Heavy Parallel Loads

View Specs 80 GB and up
Compare Benchmarks

Pay Only While It Runs

A server you keep bills around the clock. Set the hours you actually work and compare them with the same card left on all month, at today's price.

Left on all month$0
Kilnara, your schedule$0

You Save Each Month

$0

0% less than keeping it on

Rent a GPU

Short-Lived Machines for Serious Work

Leave an email for the token launch, or open the platform and start a workspace now.

CA: Launch soon · or rent a GPU now