Skip to main content
This page describes a deprecated Thunder Compute instance model. For current setup instructions, see the Quickstart guide.
Thunder Compute previously offered two instance modes: prototyping (later called development) and production. The choice affected the intended workload, price, and compatibility of an instance.
  • Prototyping (or development) was meant for research, experimentation, and short-lived development.
  • Production was better suited for long-running inference, batch training, and production workloads.
Eventually, the proprietary GPU virtualization used for Prototyping instances became stable and reliable. At this point it was applied to all Thunder Compute instances, and the distinction became meaningless.