Anthropic reports 2 hours 12 minutes of degraded performance across Claude models

Sources: Anthropic status incident, HN discussion
Summary: Anthropic opened the incident at 16:20 UTC on 2026-08-18 for degraded performance across multiple models, naming Claude Mythos 5, Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5, and others. It narrowed to elevated errors on Claude Opus 5 at 17:12 UTC, implemented a fix at 18:26 UTC, and resolved at 19:01 UTC, stating impact ran 16:11 to 18:23 UTC. No root cause was published.
Why it matters: The incident affected claude.ai, the Claude API, Claude Code, and Claude Cowork together, so agent workloads and interactive use degraded at the same time.
Follow-up: This is the second Anthropic model-serving incident in two days, after the 2026-08-17 incident carried on the previous digest. Track whether a cause is published for either.