Livenerf: Has Opus 5.5 been nerfed yet?
As of September 29, 2026, the Livenerf benchmark has collected data for 6 out of the 30 days following the launch of Anthropic's Opus 5.5 on Claude Max. This benchmark aims to detect if a frontier model's quality deteriorates post-launch, with the current focus on Opus 5.5. The benchmark runs deterministically, using a frozen prompt, pinned CLI, exact grader, and raw logs.
It measures statistical drift over thousands of samples and follows Anthropic's Adding Error Bars to Evals protocol. The benchmark started on September 24, 2026, and will continue for 30 days, with results published after day 20. The primary metric is the paired per-item score difference against the launch-week baseline, while the secondary signal tracks the output token count per sample.
The benchmark is designed to detect changes in model quality in either direction, without assuming a specific mechanism.
Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
