Vast.ai
thecontinuitylayer.com/tools/gpu-hosting/vast-aiCost-conscious tinkerers and researchers who are comfortable picking a host themselves
Marketplace · Vast.ai
Vast.ai is a marketplace where independent hosts and data centers list GPU machines and you rent them. You can rent on demand, bid for cheaper interruptible instances, or reserve a machine for a discount.
It is often among the cheapest places to run an open model on a rented GPU, but quality varies from listing to listing. Check each host's reliability score and verified status, and expect to set up your own container. Prices move with supply and demand, so check the live listings for current rates.
Specs
| Price per hour | Varies; marketplace prices for each GPU change with supply and demand |
|---|---|
| GPUs | Marketplace of consumer and data-center GPUs from independent hosts, from older RTX cards to H100 |
| Rental types | On-demand, interruptible (spot), reserved |
| RTX 4090 | Varies (marketplace) |
| H100 SXM | Varies (marketplace) |
| H100 PCIe | Varies (marketplace) |
Good for
- Low hourly prices for a 24GB card
- Batch jobs that can restart if interrupted
- Experimenting with many GPU types
Not for
- Production endpoints needing steady uptime
- Sensitive data on unverified hosts
See also
- Lambda · Teams that want plain, well-run NVIDIA machines for training or serving larger open models
- CoreWeave · Companies running open models at large scale that need whole GPU servers and clusters
- Together AI · Teams that want a dedicated, managed endpoint for an open model without running servers
We last reviewed Vast.ai on 31 October 2026. Prices and specs change; check vast.ai before you buy or sign up. Figures we could not confirm on the vendor's own page read “not published” or “varies”.
Vendor source: https://vast.ai/pricing/gpu/RTX-4090