Runpod
thecontinuitylayer.com/tools/gpu-hosting/runpodBuilders who want a cheap GPU by the hour or a serverless endpoint for an open model
GPU cloud · Runpod
Runpod rents GPUs in three main forms: Pods (a GPU machine you control), Serverless endpoints that scale up and down with traffic, and multi-node clusters. Prices are shown per hour but billed by the millisecond, and the same GPU comes in a cheaper Community Cloud tier and a pricier Secure Cloud tier run in vetted data centers.
For open models, most people start a Pod from a ready-made template for vLLM, Ollama or a text-generation web UI, or deploy a model to a Serverless endpoint. Prices change often and depend on availability, so check the live page. Community Cloud hosts vary in reliability, so use Secure Cloud for anything customer-facing.
Specs
| Price per hour | RTX 4090 from $0.34/hr; H100 PCIe from $1.99/hr (Community Cloud) |
|---|---|
| GPUs | Consumer and data-center NVIDIA GPUs, from RTX 4090 to H100 and newer Blackwell cards |
| Billing | Hourly rates, billed by the millisecond |
| RTX 4090 | $0.34/hr Community, $0.89/hr Secure |
| H100 PCIe | $1.99/hr Community, $2.89/hr Secure |
| H100 SXM | $2.69/hr Community, $3.99/hr Secure |
| Products | Pods, Serverless, Instant Clusters |
| Tiers | Community Cloud and Secure Cloud |
Good for
- Trying an open model for a few dollars
- Serverless inference that scales to zero
- Fine-tuning runs on one or a few GPUs
Not for
- Teams needing fixed long-term prices
- Workloads that cannot tolerate a Community host going offline
See also
We last reviewed Runpod on 31 October 2026. Prices and specs change; check runpod.io 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://www.runpod.io/pricing