Users Complain of Sharp Quality Decline in New Claude Models

From August 17 to 18, multiple long-time Claude users publicly complained that the newer models (Opus 5, Fable 5) have degraded noticeably.

Confirmed

@madscene, a heavy Claude Max subscriber, reported a severe decline in Opus 5 and Fable 5, including the model frequently using odd jargon like "chips" for UI elements the user never uses.
@coocoocoo8956 described the "Claudish" phenomenon (verbose, odd tone) and Opus 5 lazily splitting tasks and adding useless comments, though he stays with Claude for its productivity.
@yacineMTB relayed developer feedback that the latest Claude lost its conversational feel; some commenters hope Claude 6 returns to form.
@haydendevs mocked the generated code as slop, joking he would quit unless it could write code that would move Linus.
@SuhailKakar aggregated complaints that models (Fable and Opus) became slow with markedly worse output; some users suspect Anthropic deliberately nerfed them ahead of a new release.
@hermanschutte criticized Claude's sycophancy and tendency to correct users or insert irrelevant trivia, and is considering switching to Codex.

The claim that Anthropic deliberately weakened existing models before a new release is pure user speculation with no evidence or official response.

The wave of complaints signals reputation risk for Claude among heavy paying users, with language style, task-splitting behavior, response speed, and code quality as the main pain points.
User attitudes are split: some consider migrating to Codex, while others stay for productivity—competitive pressure and user stickiness coexist.

2026-08-17 ~ 2026-08-18 · 6 related posts