Leaderboard Shifts After Removing SimpleQA

scaling01 · x · 2026-07-19

A screenshot shows a model leaderboard where removing SimpleQA alters the scores and rankings of multiple models. For instance, GPT-5.6 Sol retains the top spot, while models like Claude, Gemini, and Kimi see varying degrees of upward or downward movement. The poster uses this to call out individuals who are "fabricating numbers" to arbitrarily downgrade ECI in order to force a specific conclusion.

Related event: Debate Over SimpleQA's Impact on Model Rankings(3 posts)→

Original post →

More from Models

Models channel →