BenchLeader

BenchLeader runs no evaluations of its own. Once a day it collects results that independent evaluators and arenas publish in machine readable form, such as Artificial Analysis, LMArena, Epoch AI, LiveBench and SWE-bench, matches the different names each source uses for the same model and reasoning setting, and ranks them together in the BenchLeader Index.

Its methodology page explains how disagreement between sources becomes an interval around each score, so the site can say how many models sit within the leader's margin of error. Scores from launch posts are kept out of the rankings, and the few vendor reported figures are labelled. The site also tracks price changes and lists the best value models among the top group.