Renting an instance on Vast.ai involves three main steps that work together to get you the exact computing environment you need.Documentation Index
Fetch the complete documentation index at: https://docs.vast.ai/llms.txt
Use this file to discover all available pages before exploring further.
The Rental Process
Choose Your Template
Select or customize a Docker template that defines your software environment - PyTorch, TensorFlow, Stable Diffusion, or custom configurations.
Find Your GPU
Search and filter through available offers to find GPUs that match your performance needs and budget.
What You’ll Need to Decide
Before renting an instance, you’ll make key decisions about:- Software Environment: Which Docker image and launch mode (SSH, Jupyter, etc.)
- Hardware Requirements: GPU type, VRAM, CPU, and system RAM needed
- Storage Size: How much disk space (cannot be changed after creation)
- Instance Type: On-demand for reliability or interruptible for cost savings
- Budget: Maximum price you’re willing to pay per hour
Quick Start Path
For the fastest path to a running instance:- Use a Recommended Template - Pre-configured and tested
- Sort by Price - Find the best deals quickly
- Check Reliability Score - 95%+ for important work
- Start with On-demand - Upgrade to reserved later if needed
What’s in This Section
Instance Types
Understand on-demand vs interruptible vs reserved options
Choosing a Template
Understanding and selecting templates for your needs
Finding & Renting
Search, filter, configure, and rent GPU instances
Reserved Instances
Save up to 50% with prepayment options
Common Questions
How long does it take to start an instance?
Usually under 5 minutes with recommended templates. Large custom images may take longer on first launch.Can I change my configuration later?
- Template: No, you need a new instance
- Disk size: No, fixed at creation
- Instance type: Can convert on-demand to reserved only