What Runpod does
Runpod provides GPU Pods, autoscaling Serverless endpoints, storage, templates, and clusters for AI development, training, and inference.
Runpod is a GPU-focused cloud. Pods provide dedicated containers, Serverless runs autoscaling handlers, and clusters support larger distributed work. Customers provide and operate the model and application code.
Rates vary by GPU, cloud type, product, region, and availability. Pods are billed by the second, storage is separate, and savings plans require prepaid commitments. Serverless worker type and activity affect cost; verify the live console.
Stopped Pods can keep charging for persistent storage, while temporary disks can be erased. Use trusted images, scoped secrets, endpoint authentication, backups, spend alerts, and termination checks. Benchmark cold starts and capacity availability.
How Runpod works
Users select GPU capacity, a container, storage, networking, and runtime controls for a Pod, or package a handler and image for Serverless workers. Runpod schedules and meters the workload, exposes configured endpoints, and can scale workers with queue demand. Developers secure images and endpoints, persist artifacts, monitor jobs, and validate model output.
How to set up Runpod
Choose the product
Select Pods, Serverless, or clusters from persistence, latency, and concurrency needs.
Estimate total cost
Check live compute, worker, storage, and commitment rates.
Build a trusted image
Pin and scan dependencies, models, handlers, and base images.
Secure the workload
Use scoped secrets, authentication, restricted ports, backups, and rate limits.
Test and monitor
Measure cold starts, throughput, memory, queues, failures, quality, and spend.
Runpod FAQs
Is there one GPU price?
No. Rates vary by accelerator, region, capacity, availability, and product.
Pods or Serverless?
Pods are dedicated containers; Serverless runs packaged workers that scale with demand.
Do stopped Pods cost money?
Compute stops, but persistent volume storage may keep billing.
Are savings plans refundable?
Current documentation describes prepaid three- or six-month plans as non-refundable.
Does Runpod validate models?
No. Customers own model quality, safety, inputs, outputs, and application behavior.
Listing reviewed 2026-08-01. Product details and pricing can change; verify important terms on the provider's website.
Related Coding AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.