2026-10-27 · America/Los_Angeles · 社区动态 · #11

Artificial Analysis Intelligence Index v5 (major update)

Artificial Analysis Intelligence Index has been updated to v5, which according to them is a major upgrade resulting in a more accurate ranking. Notably, this update widens the gap between open-weight models and closed ones. Some people already questioning its accuracy based on their experiences with models (GPT-Terra-6.1 > Fable-5.1). Also, this update, v5, has 60% of the Index based on evaluations with private test sets or held-out answers, which makes reproducing the results impossible. From their announcement: “The open weights gap widens from 3 points to 12: Kimi K3.1 Pro (weights not…

热度 46.6 / 100；排名与评分保留该期记录。