AI Glossary

Cold start

What does “Cold start” mean?

Cold start: Extra delay before a resource that is not already running can handle a request.

Cold start visual guide: Extra delay before a resource that is not already running can handle a request.
Visual guide Open full size (new tab)

What is an example of “Cold start”?

A server must load model weights before answering its first request after idling.

Keep learning

See also

Model servingTermInference serverTermModel compressionTermInference queueTermAutoscalingTerm

Same topic

Model registryTermWarm replicaTermCompute clusterTermCloud AITermCompute-bound workloadTerm