AA‑Briefcase benchmark exposes LLM efficiency gaps on real work tasks
AA‑Briefcase is a new benchmark that evaluates large language models on 11 real‑world knowledge‑work tasks. It reveals that longer runtimes and more interaction turns don’t guarantee higher scores, exposing efficiency gaps across leading models. The suite gives product teams a concrete way to compare LLMs on work‑like performance.