1. What is RunPod?
RunPod is a GPU cloud platform focused on AI scenarios, providing two core services: on-demand GPU cloud instances (ideal for training and debugging) and serverless GPU inference (ideal for publishing models as APIs). The per-second billing model allows individual developers to access compute power at a lower cost.
2. Core Features
- Elastic GPU Instances: Rent various GPUs billed by the hour or second.
- Serverless Inference: Deploy models as auto-scaling APIs.
- Pre-built Templates: One-click deployment for common AI applications (e.g., ComfyUI, Whisper).
- Network Storage: Persistent data sharing across instances.
3. Who It's For
- Developers needing elastic GPU compute to train or fine-tune models.
- Startup teams deploying AI inference services.
4. FAQ
How does RunPod billing work? You are billed by GPU usage duration (in seconds), with different prices for different GPUs. Billing stops when you stop using it, so there are no idle costs.