论文
40AI 评分(两次独立评分的平均)

Parallel Search 搜索索引速度与得分双优

原标题:Time per Task includes search time and model time. Parallel Search (fast, objective) takes 15.5s per task, among the fastest on the board...

Artificial Analysis@ArtificialAnlys发布时间 (北京时间)

Artificial Analysis 分析显示,Parallel Search 快速模式每任务耗时 15.5s,与 Octen(15.4s)、Keenable Search(实时,15.8s)相当,远低于榜单中位数约 28s,其中搜索仅占 3.8s。

中文译文尚未完成,先为你展示原文。

Time per Task includes search time and model time. Parallel Search (fast, objective) takes 15.5s per task, among the fastest on the board and comparable to Octen (15.4s) and Keenable Search (realtime) (15.8s), against a board median of about 28s.

Search takes 3.8s of that; the rest is model time. Parallel Search (advanced, objective) takes 24.2s per task (11th of 26), of which 12.7s is search: each advanced search returns in about 2.7s, against about 0.8s for fast.

On Artificial Analysis Search Index vs. Time per Task, both modes are on the frontier: no product scores higher in less time.

评测基准现象与趋势#评测/基准#搜索#现象/趋势

同一件事的其他报道3 篇