Interactive tool
GPU price explorer: compare rental cost per GPU-hour
The GPU price explorer compares cloud GPU rental cost per GPU-hour across 174 instance rows, 13 providers, and 23 GPU models. Filter by GPU model, provider, region, and on-demand versus spot, cap the price, and sort by per-GPU rate. Per-GPU rate is the instance hourly price divided by GPU count, so a single card compares fairly with a multi-GPU box. Rates are rental list prices, not card MSRP, and exclude committed-use discounts. For a curated overview of the cheapest rate per GPU, see the GPU pricing guide.
GPU pricing guide - Cheapest rate per GPU at a glance →
Example: cheapest on-demand rates in this catalog
A static snapshot of the cheapest tracked on-demand row for eight GPU models. Filter the explorer for the full catalog.
| GPU model | Per-GPU/hr | Provider | Region |
|---|---|---|---|
| V100 | $0.17 | TensorDock | published |
| L40 | $0.20 | Massed Compute | published |
| H100-NVL | $0.36 | Massed Compute | published |
| RTXPRO6000 | $0.55 | Massed Compute | published |
| RTX6000-Ada | $0.57 | Massed Compute | published |
| T4 | $0.59 | Modal | published |
| L4 | $0.80 | Modal | published |
| L40S | $0.97 | Massed Compute | published |
What you can do
- Search any GPU, instance type, provider, or region.
- Filter to one GPU model, or toggle providers on and off.
- Switch between on-demand and spot rates (spot is cheaper but interruptible).
- Cap the maximum price per GPU-hour and sort by the cheapest rate.
- Open any provider or GPU for the full per-region breakdown.
GPU price explorer. ByteCosts. Updated September 4, 2026. https://bytecosts.com/gpu/explorer/