Provider pricing

RunPod AI API pricing

RunPod offers two GPU products: Serverless for autoscaling per-second inference of LLMs and image models, and Pods for on-demand or spot GPU instances. The Serverless catalog includes vLLM-powered Llama 4, Qwen3, and DeepSeek endpoints, plus FLUX and Stable Diffusion at some of the lowest GPU-hour rates on the market.

Founded2022
HeadquartersWilmington, DE
Tracked modelsDirectory

Best for

Cost-sensitive teams serving custom models or running large batch GPU jobs.

Strengths

  • Among the cheapest H100/H200 hourly rates on the market
  • Serverless vLLM endpoints for any HuggingFace model in one click
  • Per-second billing with sub-200ms cold starts on warm workers
  • Spot instances for batch jobs at deep discounts