Compute
GPU marketplace & spot
Peer-to-peer and interruptible GPUs at the lowest prices.
What it is
Marketplaces aggregate GPUs from many data centres and individual hosts, and spot instances sell spare capacity cheaply. Prices are the lowest available, but machines can be interrupted or vary in reliability.
How pricing works
Per GPU-hour, set by hosts or by a bidding market; prices change constantly. Shown prices are a recent median.
We compare everything in one unit: USD per GPU-hour.
Typical uses
- Experiments and hobby projects
- Fault-tolerant batch jobs
- Short fine-tuning runs
What to compare
- Price vs on-demand
- Host reliability score
- Interruption risk
- Network speed
Watch out for
- Spot/interruptible machines can disappear mid-job; checkpoint frequently.
- Community hosts may not meet compliance requirements.
All 8 APIs, cheapest first
Free to browseTick 2 to 4 APIs to compare them side by side.
| API | $ / GPU-hr | GPU | VRAM | Interruptible | Quality | Price data | Compare |
|---|---|---|---|---|---|---|---|
| RunPod RTX 4090 (Community) RunPod | $0.34 | RTX 4090 | 24 GB | No | 62 | Unverified sample | |
| Vast.ai RTX 4090 (median) Vast.ai | $0.35 | RTX 4090 | 24 GB | No | 60 | Unverified sample | |
| TensorDock RTX 4090 TensorDock | $0.40 | RTX 4090 | 24 GB | No | – | Unverified sample | |
| Nebius L40S (preemptible) Nebius | $0.74 | L40S | 48 GB | Yes | – | Verified Oct 6, 2026 | |
| Hyperbolic H100 Hyperbolic | $1.49 | H100 SXM | 80 GB | No | – | Unverified sample | |
| RunPod H100 (spot) RunPod | $1.50 | H100 PCIe | 80 GB | Yes | 75 | Unverified sample | |
| Vast.ai H100 (median) Vast.ai | $1.90 | H100 SXM | 80 GB | No | 80 | Unverified sample | |
| DigitalOcean MI300X (spot) DigitalOcean (Paperspace) | $2.46 | MI300X | 192 GB | Yes | – | Verified Oct 6, 2026 |