GPU cloud platforms compared: prices and price hikes
Where you rent GPUs and run AI models: GPU instances, serverless GPUs, hosted model endpoints and code sandboxes for AI agents, compared on what the same workloads cost each month.
Prices at 100 H100 hours a month
- Runpod: $199/mo · last hike Sep 2026
- Modal: $434.93/mo · last hike Oct 2026
- Hugging Face: $450/mo
- Replicate: $549/mo
- E2B: not offered · last hike Jan 2026
Prices are USD per month at on-demand, non-interruptible rates for three fixed workloads: 100 hours on one H100 80 GB GPU, one 24 GB-class GPU serving a model 24/7 (730 hours), and 1,000 hours of 2 vCPU / 4 GB agent sandboxes. They include required plan fees and separately billed CPU and RAM, net of recurring monthly credits; a dash means the product does not do that job.