A repost claims Anthropic’s Opus 5 regresses badly despite benchmark gains

rickasaurus · x · 2026-07-28

A repost of an article argues that Anthropic’s Opus 5 feels like a bad model after a few hours of use.

The author claims the model may have improved in some areas, but says its benchmark results look like “fraud” and that several behaviors have regressed badly in practice.

Original post →

More from Models

Models channel →