Was Opus 5.5 Nerfed? Livenerf Day 13 Update

A user created "LiveNerf" to independently track Opus 5.5's performance daily, investigating claims of models being "nerfed" post-release. After 13 days, the results show a baseline of 58.4% on LiveNerf, with a deviation of 6.6 points per 10-day window, which was less than the expected 7.5. The creator emphasizes the importance of tools like LiveNerf for measuring AI performance when companies lack transparency about their products.

This report provides the first independent, daily tracking of Opus 5.5's performance, showing a smaller deviation than initially expected.