Anthropic resolves Claude models and API outage
Anthropic identified and resolved a service disruption that caused elevated error rates across several models—including Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5—along with core services including claude.ai, the Claude API, Claude Code, and Claude Cowork. The incident lasted from 00:50 UTC to 02:10 UTC on September 22, 2026, with engineers identifying the root cause within 27 minutes and restoring normal success rates across all affected endpoints within approximately an hour and twenty minutes.
As agentic workflows and developer coding tools increasingly rely on direct LLM execution, infrastructure outages immediately bottleneck entire engineering pipelines rather than just web chat.
- –Tooling-wide disruption: The outage simultaneously affected claude.ai, the developer API, and terminal tools like Claude Code, highlighting the operational fragility of agent-centric workflows during provider hiccups.
- –Heterogeneous model impact: Errors selectively hit frontier models such as Mythos 5.1, Fable 5.1, and Opus 5, reflecting the operational complexity of routing across multi-tier model clusters under high demand.
- –Fast recovery time: Anthropic diagnosed the root cause in under half an hour and brought success rates back to baseline within 80 minutes, minimizing protracted downtime for enterprise and developer users.
DISCOVERED
2h ago
2026-09-22
PUBLISHED
5h ago
2026-09-22
RELEVANCE
AUTHOR
corvad