Hosts set their own prices, so rates vary by GPU and availability. You are billed for the time you hold the machine — stop early and the unused prepay is refunded.
| GPU | VRAM | Petabyte | Cloud reference | You save | Region | |
|---|---|---|---|---|---|---|
| Loading live prices… | ||||||
Prices are set by individual hosts and change with demand and availability. "Cloud reference" is an on-demand hyperscaler rate for a comparable GPU class, used as a benchmark — not a quote from any specific provider.
Every GPU's reference $/hr is derived from its FP16 TFLOPS benchmark, so a slower card is never priced above a faster one. "Live avg" is the mean of hosts currently online for that model.
| GPU | FP16 TFLOPS | Reference $/hr | Live avg $/hr | Cloud ref | Save |
|---|---|---|---|---|---|
| Loading catalog… | |||||
You prepay for a window. When you stop, we bill the hours you held (minimum one) and refund the rest to your wallet. Extend at any time.
Taken from the rental, not added on top. Hosts see their exact payout before they list.
Earnings accrue per completed rental and can be withdrawn on demand or on a weekly schedule.