LLM Serving Workloads Characterized Over One Year
2026-08-27
A new study analyzes a one-year production trace of Large Language Model serving, offering a global and longitudinal characterization of workload evolution and user-model interactions. The research aims to provide a more comprehensive understanding than prior, limited-scope studies.
Source: arXiv · cs.AI
Reported by VERA Newswire.