AI shrinkflation: Why Anthropic’s Claude Opus 4.7 may be less capable than the model it replaced

Some users say Anthropic’s Claude Opus 4.7 feels worse than earlier Claude versions, especially for reasoning, coding, and proof tasks.

Complaints on Reddit and elsewhere mention more self-corrections, weaker problem-solving, and reduced usefulness for practical developer work.

Commentators dubbed the issue “AI shrinkflation,” suggesting tighter safety and alignment may be making premium models feel downgraded.

The backlash raises concerns that vendors may be trading raw capability for safer behavior, hurting trust in paid model upgrades.