RunPod vs Vast.ai

Same workloads, both price lists, refreshed daily. On shared line items today: RunPod undercuts on 3 of 19 shared GPUs, Vast.ai on 16.

RunPod

GPU cloud with per-second billing across Secure Cloud (T3/T4 datacenters) and cheaper Community Cloud, plus serverless GPU endpoints with autoscaling. One-click templates for vLLM, Ollama and ComfyUI.

  • Per-second billing
  • Community Cloud undercuts most on-demand pricing
  • Serverless endpoints with cold-start optimization

Visit RunPod

Vast.ai

Peer-to-peer GPU marketplace: the cheapest raw $/hr anywhere. Hosts range from verified datacenters to hobbyist rigs; you trade some reliability guarantees for 2-5x lower prices than traditional clouds.

  • Lowest $/hr in the market
  • Huge GPU variety incl. consumer cards
  • Interruptible (spot) pricing for batch jobs

Visit Vast.ai

Shared GPUs, priced by both

Quotes refreshed 2026-08-28 ยท cheapest tier each
GPURunPod $/hrVast.ai $/hrCheaper
A6000 $0.330community $0.281marketplace Vast.ai
6000 Ada $0.740community $0.388marketplace Vast.ai
L40S $0.790community $0.401marketplace Vast.ai
RTX PRO 6000 $0.500secure $0.803marketplace RunPod
A100 SXM 80GB $1.39community $0.667marketplace Vast.ai
H100 SXM $2.69community $1.34marketplace Vast.ai
H200 $3.59community $3.16marketplace Vast.ai
B200 $5.98community $5.31marketplace Vast.ai
B300 $6.94community $5.63marketplace Vast.ai
V100 $0.190community $0.027marketplace Vast.ai
A100 SXM 40GB $1.00community $0.536marketplace Vast.ai
RTX 3090 $0.220community $0.081marketplace Vast.ai
RTX 4090 $0.340community $0.120marketplace Vast.ai
RTX 5080 $0.390community $0.134marketplace Vast.ai
RTX 5090 $0.690community $0.321marketplace Vast.ai
A100 PCIe $1.19community $0.454marketplace Vast.ai
H200 NVL $0.500community $2.07marketplace RunPod
H100 PCIe $1.99community $1.33marketplace Vast.ai
H100 NVL $2.59community $2.60marketplace RunPod

Cheaper-on-count is a tally of listed prices, not a quality verdict: throughput, reliability and quantization differ between providers. Disclosure.