Home / Technology / Users Claim Anthropic Model Performance Dropped
Users Claim Anthropic Model Performance Dropped
13 Apr
Summary
- Users report Claude Opus 4.6 and Code feel less capable.
- Allegations of AI "shrinkflation" and performance throttling surface.
- Anthropic acknowledges usage limit changes, denies model degradation.
Numerous developers and AI power users are voicing concerns on social media platforms, alleging a noticeable decline in the performance of Anthropic's Claude Opus 4.6 and Claude Code models. These users report issues such as decreased reliability, more frequent errors, and a perceived reduction in reasoning capabilities, with some labeling the situation "AI shrinkflation."
These complaints have gained traction on GitHub, X, and Reddit, with detailed analyses from senior AI professionals pointing to observable regressions in model behavior. While Anthropic acknowledges recent adjustments to usage limits and reasoning defaults, the company officially denies intentionally degrading its models to manage capacity. The company stated that changes to interface elements and default effort levels are the primary drivers behind user perceptions, rather than a reduction in core model quality.
