Wonder if the "5 feels degraded" complaints are actually a context-length problem, not a model problem

Opus 5 does 1M tokens. 4.6 caps at 200k. People switching to 5 for long tasks are now running way bigger contexts than they ever ran on 4.6. Bigger context, more room to lose the thread, more stale stuff dragging on attention, more cost. Anyone tested this