A rate attached to the rental.
The hourly rate is recorded with the workload, giving the usage calculation a defined starting point.
The AI resource & compute marketplace
AI compute,
on your terms.
Find verified GPU providers to train and run your models. Bring your data, choose your hardware, and manage your workloads in one place.
Connect your existing Hugging Face projects.
THE IDEA BEHIND GARBINAIGreat AI needs more
Great AI needs more
than a great model.
It needs useful data, available hardware, and people who bring them together. GarbinAI connects that chain, from the repository you start with to the provider who runs your workload.
CreatorsModels & datasets
ProvidersGPU capacity
BuildersTraining & inference
01 / THE COMPUTE MARKETPLACEThe hardware behind
The hardware behind
your next workload.
Compare the things that matter: GPU memory, region, capacity, and provider pricing. Choose your hardware with context.
garbin / gpu-marketplacePLATFORM WALKTHROUGH
COMPUTE DIRECTORY
Find your next GPU.
Region GPU model MemoryTraining + inference
Provider-listed GPU
Model, GPU count & available VRAM
Hardware verificationLOCATIONChoose your region
CAPACITYCheck available slots
PRICINGCompare hourly rates
BUY COMPUTE, WITH CONTEXTKnow the machine.
Know the machine.
Know the terms.
Find a server that fits your workload instead of shaping your workload around whatever is available.
- Hardware that is checkedLive verification compares the detected GPU model and count with the listing.
- Resources you can compareReview memory, region, workload support, and provider rates before renting.
- A choice of how to allocateSelect your provider or let compatible resources be matched to your job.
02 / THE RESOURCE LAYERCompute is only
Compute is only
part of the story.
Your models, data, and demos deserve a home too. Keep the resources behind your AI work connected.
Start with a model. Keep its context.
Discover models by task, inspect repository files, and review documentation before you commit to a foundation. Import your own Hugging Face repository when you already have one.
Browse modelsmodel / README.md main
01
# Your model, documented02
Task Text generation03
Source Hugging Face repository04
Access Defined by the owner
Your work, connected.03 / FROM RESOURCE TO RUNA job is more than
Create your first jobA job is more than
a start button.
STAGE 01
Tell us what you want to run.
Choose training or inference. Add your repository or container reference, run command, resource requirements, and duration limit.
- Workload
- Training / inference
- Source
- Repository or container
- Resources
- GPU, memory, region
- Duration
- Your chosen limit
THE OTHER SIDE OF THE MARKETPLACEYOUR HARDWARE.Someone else's
next breakthrough.
next breakthrough.
04 / FOR GPU PROVIDERSPut your available
Put your available
compute to work.
Connect the hardware you own to people who need it. Register your server, verify its GPU resources, and choose the rates and capacity you want to offer.
- 01Connect & verify
Register your machine and verify its detected GPU resources.
- 02Set your terms
Choose pricing, supported workloads, and maximum concurrent jobs.
- 03Manage your capacity
Pause or resume your listing, recheck health, and follow assigned work.
05 / FOLLOW THE VALUEThe work gets done.
The work gets done.
The usage gets recorded.
A marketplace needs a clear record of what ran, what it cost, and how the amount was split.
ANATOMY OF A JOB RECEIPTILLUSTRATIVE
GPU-hours×Agreed rate=Usage charge
Provider amountHardware contribution
Platform feeMarketplace operation
Connected to the rental, provider, and job outcome.
A breakdown you can follow.
Settlement records keep billed GPU-hours, charges, fees, and the provider amount together.
An outcome that matters.
Completed, failed, and cancelled jobs have outcomes that are reflected in the usage and settlement flow.
06 / TRUST IS PART OF THE WORKFLOWClear ownership.
Clear ownership.
Checks before compute.
Your server stays yours.
Server management is tied to the creator. Providers control their own pricing, availability, and configuration.
A listing is not a health check.
Hardware verification, server state, fresh health information, and capacity inform rental readiness.
Actions have boundaries.
Authenticated workflows and resource ownership checks govern access to jobs and provider actions.
07 / A MARKETPLACE WITH PEOPLE IN ITPublish something useful.
Publish something useful.
Build on someone else's work.
Models, datasets, and Spaces are more valuable with context and conversation. Share repositories, explore community projects, and bring collaborators into the work.
Explore the communityModel creatorsPublish the foundation
Dataset ownersShare the source
GPU providersSupply the compute
AI buildersTurn it into a product
THE LONGER VIEWAI should create value
AI should create value
for the people behind it.
The marketplace is the starting point. Our broader direction connects useful AI infrastructure with privacy and recognition for the people who contribute.
Garbin-Shield
Exploring client-side privacy tooling, sensitive-data masking, and provenance signals around AI interactions.
Contributor rewards
A longer-term model for recognizing the people who contribute models, data, training, code, and compute.
BEFORE YOU GET STARTEDA few practical
A few practical
questions.
What is GarbinAI?
GarbinAI connects AI resources and GPU compute in one marketplace. Builders can discover models and datasets, run training or inference workloads, and share demos in Spaces. Providers can register hardware, set rates, and manage their servers.
How is provider hardware verified?
Server onboarding checks the actual machine over SSH and compares the detected GPU model and count with the provider's listing. Server health and available capacity are checked as part of rental readiness.
Can I choose my own GPU provider?
Yes. Compare provider listings and select a server for your workload. Job creation also supports automatic selection for compatible resources with available capacity.
How does usage billing work?
Usage is recorded in GPU-hours using the rental's hourly rate. Settlement records track the charge, platform fee, provider amount, and job outcome. Review the current listing and billing details before starting a job.
Can I bring my existing Hugging Face projects?
Yes. Sign in with Hugging Face to import models, datasets, and Spaces. GarbinAI also supports creating repositories and managing their files and documentation.
Are the privacy and contributor reward features available?
The broader GarbinAI roadmap includes Garbin-Shield privacy tooling and contributor rewards. These are a longer-term direction, separate from the current model, dataset, Space, and GPU marketplace workflows.
CHOOSE YOUR SIDE OF THE MARKETPLACEBring the workload.