# RunPod Pricing

**Source:** https://www.runpod.io/pricing — **Snapshot:** 2026-08-12 (extracted from the page's structured data)
**Billing:** per-second, pay-as-you-go. Two pools: **Community Cloud** (cheap, no SLA, spot-reclaimable) and **Secure Cloud** (dedicated, guaranteed).

| GPU | Community $/hr | Secure $/hr |
| --- | ---: | ---: |
| RTX A5000 | 0.16 | 0.27 |
| RTX 3090 | 0.22 | 0.46 |
| RTX A6000 | 0.33 | 0.49 |
| RTX 4090 | 0.34 | 0.69 |
| RTX PRO 4500 32GB | 1.15 | — |
| L4 | 0.44 | 0.39 |
| RTX 5090 | 0.69 | 0.99 |
| L40 | 0.69 | 0.82 |
| RTX 6000 Ada | 0.74 | 0.77 |
| L40S | 0.79 | 0.99 |
| A100 PCIe | 1.19 | 1.39 |
| A100 SXM | 1.39 | 1.49 |
| RTX Pro 6000 | 1.69 | 1.99 |
| H100 PCIe | 1.99 | 2.89 |
| H100 NVL | 2.59 | 3.19 |
| H100 SXM | 2.69 | 2.99 |
| H200 | 3.59 | 4.39 |
| B200 | 5.89 | 5.89 |
| B300 | 6.94 | 7.39 |

Notes:
- Pod templates (one-click ComfyUI etc.) are the fastest path — a template adds nothing to the GPU rate but saves the entire install step.
- Community vs Secure is ~1.5–2x on most cards. Community suits retryable queues; Secure suits anything you can't afford to lose mid-job.
- A40, RTX 4000/2000 families, MIG slices also exist on the page — the table above is the video-workload-relevant subset.

**Verify live before committing spend** — this is a snapshot; rates move (especially community pool spot pricing).