Is Opus 5.5 nerfed? New benchmark called LiveNerf measures this live

A user has been tracking Opus 5.5's performance since its release, re-running independent benchmarks like GPQA and SWE-bench daily. They have not detected any "nerf" in its capabilities so far. A new benchmark called LiveNerf, which uses Claude to identify difficult questions for Opus 5.5 and compares it against Opus 5, is also being used. A deviation of more than 7.5 points on LiveNerf would indicate a potential decrease in capability. The user plans to continue monitoring and will share results if a nerf is detected.

This report introduces LiveNerf, a new benchmark that uses Claude to identify difficult questions for Opus 5.5, unlike previous benchmarks that only tracked general performance.