Tag
The article compares Claude Fable 5.1 and 5, showing a 7.5% cost reduction for long agent loops due to a 75% discount on cache reads, which dominates billing in extended sessions.
This paper analyzes a one-year production trace from Chutes to study LLM serving workloads, revealing temporal evolution and user-model interactions to improve serving system benchmarking.