Artificial Analysis updates Intelligence Index, Kimi K3.1 Pro drops nine spots as open-closed model gap widens to 12 points

Artificial Analysis version 5 of its Intelligence Index now attributes 60 percent of scoring to private test sets, aiming to curb benchmark overfitting. Open-source model Kimi K3.1 Pro fell from fourth to ninth place, widening the gap between open and closed models from 3 to 12 points. Closed-source model Grok 4.7 also dropped 12 points in the updated rankings.