Train your own AI, keep it, and run it
A member’s graphics card trains your AI. You keep the model here, and run it: try it in your browser, call it from your app, or let an AI agent use it. You pay by the second, from your HyperAI credit.
Train
Pick a ready-made recipe (finding things in photos with YOLO, PyTorch, TensorFlow, or fine-tuning a chat model) and add your pictures or data. A member’s graphics card trains it and stops at the time and cost limits you set.
Keep
Keep the model a training job made, or upload one you trained somewhere else: ONNX, PyTorch, or a chat model’s LoRA adapter. It is kept encrypted until you delete it.
Run
Try it here on a photo, or chat with your own chat model. Call it from your app with an API key, or let an AI agent such as Claude Code use it through MCP. Models run on members’ graphics cards; running them on our AI chips comes later.
Like Airbnb, for graphics cards
Start with your Passport
Sign in with your free HyperCluster Passport to train, keep and run your AI. New to HyperCluster? You can make a Passport on the way in.
What it costs
Paid from your HyperAI credit, in HYPEC. Only what works is charged.
Loading…
An hour on some common graphics cards
| Graphics card | HYPEC an hour | USD an hour |
|---|
A job is charged at the rate of the card that runs it, set when it starts. Members’ cards from an RTX 3060 with 8 GB up do the work. Laptop graphics cards, and cards not on our list, are priced by their memory, never above the desktop card of the same name.