Cold capacity is not immediately usable capacity.
MODEL SERVINGADVANCED LLMOPS
Model Loading and Warmup
Large models require time and memory to load before they can safely serve traffic.
A deployment preloads weights, warms representative requests and only then joins the load balancer.
Does readiness mean 'process started' or 'model can meet SLO'?
REMEMBERReadiness must include model readiness.
No uploads · No company data · No account required