Claude Opus 5 hit elevated errors three times in its first 72 hours
Anthropic's three-day-old flagship — now default on Max, top pick on Pro — has been unstable since Thursday's launch. The timeline, scope, and how to hedge.
Copy markdown
Three outages in 72 hours
Claude Opus 5 threw elevated errors three separate times: July 26 at 09:17 UTC, then twice on July 27 (08:16 and 11:27 UTC). Each cleared within roughly an hour — the latest back to baseline by 11:47 UTC.
It's the model you just got moved onto
Opus 5 shipped Thursday, July 24 as the new default on Claude Max and the top option on Pro. These errors land on the exact model most Claude Code and heavy API users are now routed to by default.
Scoped to Opus 5, not the whole platform
Each incident named Opus 5 specifically; Sonnet and older Opus tiers stayed operational. If your Opus 5 calls failed this week, it was the model, not your code.
How to hedge until it settles
Wire a fallback model (Sonnet 5 or Opus 4.8) into your retry path, and avoid pinning long agent runs to Opus 5 alone right now. New-flagship launch weeks reliably run hot.