Anthropic Accused of Quietly Nerfing Models Weeks After Launch, Opus 5.5 Expected to Follow

iannuttall · x · 2026-09-25

A claim circulating that Opus 5.5 will be nerfed via the harness rather than weights: Anthropic's Sonnet 4, Opus 4.6 and Opus 5 all got worse weeks after launch (later admitted), via lower default thinking effort and shorter outputs to save tokens. Opus 5.5 is sold as 40% cheaper to run. User iannuttall says he hasn't touched a ChatGPT model since Opus 5.5 launched and his usage barely rises — wondering if this time no nerf happens.

Original post →

More from Models

Models channel →