Claude Status: Elevated Errors Reported Across Multiple Models

Anthropic's Claude status page reported elevated error rates on requests to several Claude models early on 22 September 2026, opening an incident at 00:57 UTC and closing it as resolved at 02:35 UTC. The disruption affected claude.ai, the Claude API (api.anthropic.com), Claude Code and Claude Cowork.

How the incident unfolded

The first notice, published at 00:57 UTC under "Investigating", said the company was looking into elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1 and Claude Opus 5, promising further information as soon as possible.

Twenty minutes later, at 01:17 UTC, the status page moved to "Identified": the cause of the elevated errors had been found and engineers were working on a fix. At 01:35 UTC an update reported that requests to Claude Fable 5 and 5.1 and to Claude Mythos 5 and 5.1 had returned to normal success rates, while work continued on the remaining errors affecting Claude Opus 5.

Recovery and closure

By 02:11 UTC the page was in "Monitoring", stating that success rates had returned to normal across the affected models and that the team was watching closely for further problems. The incident was marked resolved at 02:35 UTC. According to the incident report, impact occurred from 5:50pm PT to 7:10pm PT — 00:50 to 02:10 UTC — roughly one hour and twenty minutes.

What it means

For developers and teams building on the Claude API, Claude Code or Cowork, the episode was short: success rates recovered within the hour, and the status page moved through the full sequence from investigating to resolved in under two hours. The report lists claude.ai, the API, Claude Code and Claude Cowork as affected services, but does not say which customers or regions were hit hardest.