AIHacker Newsabout 2 hours ago
Benchmark Tracks Opus 5.5 Post-Launch Nerfing
11 min read
Livenerf, an append-only benchmark, began daily testing of Claude Opus 5.5 on 2026-09-24, about 2.5 days after its 2026-09-22 launch. It runs 90 samples daily for 30 days to detect accuracy drops of roughly 7.5 points per 10-day window, addressing reports that Anthropic quietly degrades models after release.
Level
Hype check
Tap to vote and see what everyone thinks.
