scaling01 mocks benchmark tables: leading every eval doesn't mean the better model

scaling01 · x · 2026-11-19

scaling01 quotes the Gemini 4 Neon benchmark table with a sarcastic caption: "Neon outscores Opus 5.5 and Sol on basically every reported eval—surely it's a better model, right?" A jab at the gap between leaderboard scores and real-world experience.

Original post →

More from Models

Models channel →