Opus 5.5’s coding performance in a nearly 24-hour trial

After nearly 24 hours testing Opus 5.5 on real repositories, including overnight runs, a user says they reactivated Claude Max. They describe performance at Fable 5.1’s level on much of the work tested, alongside stronger bug discovery and light usage at both medium and high reasoning settings. The user also cites 40% lower typical running costs versus Opus 5 and 60% cheaper cache reads. They argue that cheaper cache reads may matter more than headline token prices for long Claude Code sessions because agents repeatedly reuse context.