RunPod

The Cloud Built for AI.

RunPod offers on-demand and spot GPU instances across a global network of data centers. You pick a GPU type (from consumer-grade to A100s and H100s), deploy a container, and pay by the hour. Their serverless platform lets you deploy models as auto-scaling API endpoints without managing infrastructure.

The platform supports both interactive development (GPU pods with SSH and Jupyter access) and production workloads (serverless endpoints with automatic scaling to zero). Pricing is competitive, especially on spot/community instances, making it a popular choice for teams that need GPU compute without long-term commitments.

Pricing: Hourly

Hosting Cloud

Pricing Usage Based, from $0.06/hr

HQ 🇺🇸 United States

Founded 2022

License PROPRIETARY

Compliance SOC 2 · HIPAA · GDPR

Visit website →

Pricing

Posts

Suggest changes

Resources

Deploy and Use any Open Source LLMs using RunPod

RunPod Stable Diffusion: Getting Started

What is RunPod?

RunPod is a GPU cloud platform built for AI and machine learning workloads. It provides on-demand and spot GPU instances across a global network of data centers, along with a serverless platform for deploying models as auto-scaling API endpoints.

GPU Pods

GPU Pods are RunPod's core offering. You select a GPU type (A40, A100, H100, and others), choose a pre-built template or bring your own Docker container, and get a full Linux environment with SSH and Jupyter access. Storage is handled through network volumes that persist across pod restarts. On-demand pods run at fixed hourly rates, while community/spot instances offer the same hardware at lower prices with the trade-off of potential interruption.

Serverless Platform

RunPod Serverless lets you deploy models as API endpoints that scale automatically based on incoming requests. You package your model in a Docker container with a handler function, and RunPod manages the infrastructure. Endpoints can scale to zero when idle (you only pay for compute time) and scale up to handle bursts. Cold start times depend on model size and container setup.

Pricing

RunPod uses hourly pricing for GPU pods, with rates varying by GPU type. Community cloud instances are cheaper than secure cloud instances. Serverless is billed per second of compute time. There are no minimum commitments or long-term contracts. Volume storage is billed separately per GB per month.

Who should use RunPod?

RunPod works well for teams that need flexible GPU access without managing their own hardware. It's popular for model training, fine-tuning, and inference workloads. The serverless platform is a good fit for deploying custom models that need to scale with traffic without paying for idle GPUs.