Skip to main content
Vast.ai operates as a marketplace, connecting users with GPU providers worldwide. This means prices vary based on supply and demand, giving you access to competitive rates for GPU computing.

​
How Vast.ai Pricing Works

Unlike traditional cloud providers with fixed pricing, Vast.ai uses a marketplace model where hosts set their own prices. This creates competitive rates without static price quotes—the market determines pricing in real-time. You can easily check current market rates:
Serverless Pricing: Vast Serverless auto-scales instances with no separate pricing tier—you pay only for the underlying instance costs (compute, storage, bandwidth) with no additional fees. See Serverless Pricing for details.

​
Key Pricing Principles

  • Market-driven rates: Prices fluctuate based on real-time supply and demand
  • Combined pricing: Your total cost includes a combination of GPU compute, storage, and bandwidth charges
  • Pay for what you use: Billed by the second for actual usage
  • Service level pricing: Costs vary based on dedicated or interruptible access and term length

​
What does Vast cost compared to other providers?

Vast typically delivers more competitive pricing than traditional cloud providers. The marketplace model creates natural price competition—hosts compete for your business rather than locking you into fixed rates. Key advantages:
  • Direct competition between hosts drives rates lower
  • Real-time supply and demand optimizes pricing
  • No markup layered on top of host-set prices
  • Global GPU supply improves availability and cost
Check current market rates in the Dashboard or query programmatically via the CLI or API.

​
Cost Components

Your total rental cost includes three primary components:

​
GPU Compute

Charged per second while your instance is running. Rates vary significantly based on:
  • GPU model: High-end models (H100, A100) command premium rates
  • GPU quantity: Multi-GPU configurations increase costs proportionally
  • Host reliability: Higher reliability scores typically correlate with higher prices
  • Geographic location: Regional supply and demand affects pricing
  • Market conditions: Real-time marketplace dynamics influence rates

​
Storage

Billed continuously while your instance exists, regardless of running state. Storage costs vary by host and are typically higher for stopped instances than running instances.
Storage charges continue even when instances are stopped. Delete instances completely to cease storage billing.

​
Bandwidth

Data transfer costs vary by host and include both upload and download traffic. Charges apply per byte transferred. Review bandwidth rates during instance selection as these can significantly impact total costs for data-intensive workloads.

​
Rental Types

Vast.ai offers three rental options, each suited for different workloads and budgets:

​
On-demand

Best for: Production workloads, time-sensitive tasks, guaranteed availability
  • Fixed price set by the host
  • Price locked for your entire rental period
  • Runs continuously until you stop it
  • Your instance won’t be stopped by other users

​
Reserved

Best for: Long-term projects with predictable resource needs
  • Convert on-demand rentals to reserved for up to 50% discount
  • Price locked for your entire reservation period
  • Prepay for extended periods in advance
  • Cannot migrate between different hosts
  • Refund protection against host failures

​
Interruptible

Best for: Batch processing, fault-tolerant workloads, cost optimization
  • Bid-based pricing system
  • Can be 50%+ cheaper than on-demand
  • May be paused if outbid
For detailed rental type comparisons, see Rental Types and Reserved Instance Discounts.

​
Billing Basics

Here are the essential billing concepts to understand:
  • Credits required upfront: You must add credits to your account before starting any instances
  • Automatic instance suspension: When your credit balance reaches zero, instances are automatically stopped but remain in your account
  • Ongoing storage charges: You continue to be billed for storage on stopped instances until they are deleted
For complete billing details, see the billing page.
⌘I