Skip to main content
AI Tool Radar

RunPod

PaidNew

GPU cloud with per-second billing for AI training, fine-tuning, and inference

4.1/5Editor's Rating
Starting at From $0.19/hrUpdated 2026-07-19

What is RunPod?

RunPod is a GPU cloud for developers who need serious compute without hyperscaler pricing: on-demand pods from $0.27/hr (RTX A5000) to $7.39/hr (B300), serverless endpoints with sub-200ms FlashBoot cold starts, and instant multi-node clusters. Billing is per second, egress is free, and one-click templates get open-source models running in minutes. Pay-as-you-go with no subscription; used by teams at Hugging Face, Perplexity, and Cursor.

Key Features

On-demand GPU pods from RTX A5000 ($0.27/hr) to B300 ($7.39/hr), billed per second
Serverless endpoints with scale-to-zero and sub-200ms FlashBoot cold starts
Secure Cloud (data-center grade) plus cheaper Community Cloud capacity
Instant Clusters for multi-node training without a sales call
Hub with one-click templates for popular open-source models and tools
Native GitHub deployments with rollback
Network volumes from $0.07/GB/mo and zero ingress/egress fees
REST API and CLI for automation, default $80/hr spend cap as guardrail

Pros & Cons

Pros

  • Among the cheapest rates anywhere for H100s and RTX 4090s, billed per second
  • No egress fees, where hyperscalers charge $0.09-0.12/GB on the way out
  • FlashBoot cold starts under 200ms make serverless inference genuinely usable
  • Huge GPU selection: roughly 20 types from consumer cards to B300
  • Pay-as-you-go with no subscription; a pod is live within minutes

Cons

  • Trustpilot 3.5/5: recurring reports of pods failing to start or crashing while billing continues
  • Stopped pods keep billing for storage; idle volume disk costs $0.20/GB/mo, double the running rate
  • GPU availability fluctuates at peak times, and the dashboard has shown unavailable GPUs as available
  • Community Cloud is cheaper but reliability varies; no SLA on standard plans
  • Spot pricing has been changed abruptly in the past (raised 25%, removed, re-gated behind the API)

Pricing

From $0.19/hr

paid model

You might also like