Opus 4.8 Tops AAI Benchmark as GPT-5.5 Matches Cost-Perf Curve

Negative users criticize Opus models like 4.5 for over-alignment that causes excessive refusals on ordinary prompts.

1 comment with sentiment. View comments.