Train your own AI on the community’s graphics cards, then call it from any app or agent.
Members share the graphics cards in their PCs, and you use them by the second: train your AI, keep it, then call it from any app with an API key, or from AI agents such as Claude Code and Cursor, through MCP. No graphics card to buy. Made for schools, students and small businesses.
How it works
- 1
Train
Upload your pictures or data and pick a ready-made recipe, such as “find things in photos”. A member’s graphics card trains your AI and stops when it’s done.
- 2
Keep
Save your finished AI. It’s kept encrypted until you delete it. Or upload one you already have (ONNX, PyTorch .pt, or a chat model’s LoRA adapter as a .zip), or add it from a link.
- 3
Try it
Drop in a photo and see what your AI finds, with boxes drawn around each thing. Or chat with a chat model you fine-tuned.
- 4
Use it anywhere
Connect it to your own app or website with an API key, or let an AI agent such as Claude Code or Cursor use it.
Two simple services
Train: a graphics card when you need one
Use a member’s graphics card for as long as your training takes, then let it go. Ready-made recipes for PyTorch, YOLO (finding things in photos), TensorFlow and fine-tuning a chat model, or bring your own.
Run: your AI, ready to answer
Your app or AI agent sends a picture, your AI answers with what it found. Or use ours: HyperCluster’s AI chips look at a picture for under half a cent.
Who it’s for
Schools and teachers
AI lessons where every student trains a real model, without a lab full of expensive computers.
Students and makers
Projects and competitions without a cloud bill that keeps running after you’ve gone to bed.
Small businesses
Read number plates at the car park gate, count stock on a shelf or read receipts, without hiring an AI team.
What it costs
You pay only while your AI works, by the second, from your HyperAI credit. No setup, no monthly bill, nothing left running by mistake. You set a time and cost limit for every training job, and it stops there.
| Graphics card | $ an hour | HYPEC an hour |
|---|---|---|
| RTX 3060 · 12 GB | $0.068 | 6.8 |
| RTX 4070 · 12 GB | $0.143 | 14.3 |
| RTX 3090 · 24 GB | $0.212 | 21.2 |
| RTX 4090 · 24 GB | $0.489 | 48.9 |
| A100 PCIE · 40 GB | $0.585 | 58.5 |
| RTX 5090 · 32 GB | $0.660 | 66.0 |
Each graphics card has its own price an hour, set by HyperCluster from what the same card rents for elsewhere. Billed by the second from your HyperAI credit.