Claude Opus 5.5 Faces Unproven 'Nerf' Claims Days After Strong Launch

Anthropic released Claude Opus 5.5 on September 22, 2026, as a flagship model offering 40% cost savings and 30% faster performance than its predecessor, but independent developers launched a real-time benchmark just one week later amid user complaints of quality degradation. The model arrived with impressive credentials: 66.4% accuracy on agentic coding tasks, pricing at $4 per million input tokens (versus $5 for Opus 5), and documented successes in real-world use—one user completed a 680,000-line code migration in under a day. Within days, however, developer communities reported that responses felt duller, more cautious, and reminiscent of older Claude models, sparking debate over whether Anthropic had deliberately reduced the model's capabilities.