Users on Reddit are reporting a significant degradation in the performance of Anthropic's Claude models, specifically Opus 5.5 and Sonnet 5.5. These users describe the models as producing "absolute trash," hallucinating frequently, and failing to follow instructions. The sentiment among these users is one of frustration and confusion regarding the perceived "nerfing" of the AI's capabilities. AI
IMPACT User reports suggest potential issues with recent Claude model updates, impacting reliability and performance.
RANK_REASON User-generated discussion on Reddit about perceived model degradation, not an official release or benchmark.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →