benchmark

【数据与研究】Use when choosing between engines or models for an agent pipeline and the answer must come from measurement on your own data, not from marketing pages: speech recognition for meeting recordings, the model that turns a transcript into notes, the critic model that checks it. Runs every variant in an isolated container, measures time, cost per hour of i…

独立 GitHub 快照 · 自动分类 · 以原作者文档为准