RunPod

What is RunPod?#

RunPod is a GPU cloud platform built for AI inference and training workloads. It provides on-demand and spot GPU instances for running models, fine-tuning, and serverless AI endpoints — with per-second billing and data centers across North America and Europe.

What can RunPod do?#

• 01

  On-Demand and Spot GPU Instances

  Rent GPU instances by the second with on-demand (guaranteed) and spot pricing across A100, H100, RTX 4090, and other AI-grade GPUs.

• 02

  Serverless GPU Endpoints

  Deploy AI model inference as serverless endpoints that autoscale to zero when idle, with per-request pricing to avoid idle charges.

• 03

  Pod Templates

  Launch pre-configured pod environments for PyTorch, ComfyUI, Stable Diffusion, Axolotl, and other AI frameworks without manual setup.

• 04

  Network Volumes

  Attach persistent network storage across pods so model weights and datasets persist between GPU sessions without re-downloading.

• 05

  vLLM and LLM Inference Support

  Deploy LLMs with optimized vLLM inference backends for production-scale language model serving with batching and streaming.

Use Cases#

• “ML engineers — Fine-tune open-source LLMs like LLaMA on large datasets using spot GPU instances to minimize compute costs during experimentation.
• “AI developers — Deploy a custom Stable Diffusion checkpoint as a serverless endpoint that autoscales based on request volume without paying for idle capacity.