Skip to main content
Runtime offers three machine tiers. All deployments use suspended state by default - machines pause when idle and resume on incoming requests with fast cold starts.

Choosing a tier

256 MB RAM • 1 shared CPUThe default for most workloads. Good for simple APIs, webhooks, and tools.

Suspended state

Runtime uses Fly.io’s suspended state instead of full machine stops:
  • Machines suspend after a period of inactivity
  • Resume in ~300-500ms on incoming requests (vs 1-2s for cold start)
  • Memory state is preserved during suspension
  • You only pay for active compute time
This gives you the cost benefits of serverless with near-instant response times.

Setting the tier

In your manifest

At deploy time

Override the manifest tier:

Change without rebuilding

Use config-only deploys to change tier instantly:

Template requirements

When to upgrade

  • Getting out-of-memory errors → upgrade tier
  • Response times degrading under load → upgrade tier
  • Running AI models or heavy processing → use performance
Check memory usage in logs: