RunPod vs Vast.ai
Both appear in 1 section: gpus. RunPod is cheaper on 4 of 6 comparable measures.
These two are the closest thing to a direct comparison of marketplace GPU rental, and the distinction matters more than the prices. Vast.ai is entirely a marketplace: every row in Pikkly's data for Vast is community tier, meaning the hardware belongs to third-party hosts who set their own prices and reliability. RunPod runs both models, a community tier with the same economics and an on-demand tier on its own infrastructure. Compared like with like, community against community, they are close: an H100 at $2.27/hr on Vast against $1.99/hr on RunPod, an H200 at $3.60 against $0.50, a B200 at $5.64 against $5.98. The spread within either marketplace at any moment is wider than the gap between them, because both are really aggregating many independent sellers. Vast's advantage is depth of supply and a genuine auction: if you are willing to filter on reliability scores and bid, it is usually possible to find capacity below anything a first-party provider charges, and its interruptible pricing goes lower still. RunPod's advantage is that the on-demand tier gives you an exit. When a job actually matters you can pay $2.89/hr for an H100 on RunPod's own hardware without changing platform, tooling or storage. If your work is all short and restartable, Vast will usually be cheaper. If some of it is not, RunPod covers both cases.
GPUs
All gpus →| Metric | RunPod | Vast.ai |
|---|---|---|
| B200per GPU-hour | $5.98community↗ | $5.64Cheapercommunity↗ |
| B300per GPU-hour | $6.94Cheapercommunity↗ | $8.75community↗ |
| H100 PCIe 80GBper GPU-hour | $1.99Cheapercommunity↗ | $2.40community↗ |
| H100 SXM 80GBper GPU-hour | $2.69community↗ | $2.27Cheapercommunity↗ |
| H200per GPU-hour | $0.500Cheapercommunity↗ | $3.60community↗ |
| RTX 3090per GPU-hour | $0.220Cheapercommunity↗ | $17.34community↗ |