Overage / GPU cloud
GPU cloud pricing comparison
7 services, one row each: what it costs, the tightest limit you will hit, and when it last changed. Sorted by price per normalized unit; click a column header to re-sort.
| Crusoe Cloud | yes | 13 modelsper-model pricing | $0.05 in$0.20 out / 1M tokens | unlisted | 2026-10-24 | none |
| Hyperstack | none listed | 4 modelsper-model pricing | $0.10 in$0.20 out / 1M tokens | unlisted | 2026-10-24 | none |
| RunPod | none listed | 21 instancesper-instance pricing | $0.27/ GPU-hour | unlisted | 2026-10-24 | none |
| Modal | 30 vendor credits / mo | 11 instancesper-instance pricing | $0.59/ GPU-hour | unlisted | 2026-10-23 | none |
| Lambda | none listed | 22 instancesper-instance pricing | $0.69/ GPU-hour | unlisted | 2026-10-24 | none |
| DigitalOcean GPU Droplets | none listed | 16 instancesper-instance pricing | $0.76/ GPU-hour | unlisted | 2026-10-24 | none |
| Nebius | none listed | 7 instancesper-instance pricing | $1.55/ GPU-hour | unlisted | 2026-10-24 | none |
Model services: the cheapest input and cheapest output price per 1M tokens across all of the service's models (not necessarily the same model; see every model on the token pages). GPU services: the cheapest instance per GPU-hour (price divided by GPUs per instance). Services with no listed rate show "unlisted" and sort last. All categories.
Compare
Every pair in this category has a page, for example Crusoe Cloud vs Hyperstack. All comparisons.