Claude Opus 5.5 tops writing benchmark at 2631 Elo, a record 307-point lead

teortaxesTex · x · 2026-09-24

Claude Opus 5.5 is the new #1 on a popular AI writing benchmark at 2631 Elo — a 307-point gap over second-place Fable (2324), the largest single jump since the leaderboard launched in June 2026. It's also the first model to clear 91/100 on the rubrics. The catch: at max effort one script takes 17 minutes and $3.43, the slowest config on the board and top-5 most expensive. Commenter teortaxes notes the model's prior writing weaknesses were always fixable — Anthropic just didn't prioritize them.

Original post →

More from Models

Models channel →