Anthropic denies Claude AI model degradation claims

Developers and power users are accusing Anthropic of intentionally degrading the performance of its Claude AI models, citing issues with reasoning, task completion, and token efficiency. While Anthropic employees deny these claims, they acknowledge recent changes to usage limits and reasoning defaults. Some users have labeled this "AI shrinkflation," suggesting they are receiving a weaker product for the same price. A senior AI director at AMD presented detailed analysis suggesting a significant drop in Claude's reasoning depth and an increase in poorer performance indicators, though Anthropic attributes these changes to product and interface adjustments rather than model downgrades.