Anthropic’s Claude AI Faces User Backlash Over Declining Performance and Quickening Quotas

Anthropic's Claude AI has come under intense scrutiny as users report significant declines in performance and faster depletion of usage quotas, according to multiple sources. Complaints center on reduced ability to follow instructions and manage complex tasks, with some users characterizing the latest versions Claude Opus 4.6 and Claude Code as less reliable and prone to increased hallucinations (Fortune, VentureBeat).

The controversy intensified following a brief outage of the Claude service reported on Monday, which heightened concerns about rising operational costs and service quality (The Register). The company recently shortened the time-to-live (TTL) for the Claude Code prompt cache from one hour to five minutes, a move intended to optimize performance but which users say is causing their quotas to drain faster during long sessions (The Register).

Social media reports and benchmark comparisons have fueled speculation that Anthropic may be "nerfing" its models to control computational expenses, though company representatives have denied intentionally degrading model capabilities (VentureBeat). A viral BridgeBench post claimed a notable drop in hallucination accuracy in Claude Opus 4.6, but critics of the benchmark methodology have questioned the validity of that assertion (BeInCrypto).

As Anthropic prepares for a potential IPO, these developments raise critical questions about model tuning practices, service transparency, and user trust. How the company addresses these performance and quota concerns will be pivotal to maintaining its standing in the competitive AI landscape. Industry observers will be watching closely for upcoming statements from Anthropic and any further changes to model configurations or pricing structures.