Claude Service Incident Report: Elevated Error Rates Across Multiple Models

On June 16, 2026, Anthropic experienced a service disruption that resulted in elevated error rates for several Claude models, affecting both the web interface and API integrations. The incident occurred in two distinct phases over approximately two hours before being fully resolved.

Incident Timeline and Impact

Anthropic reported a total service disruption window starting at 17:23 UTC and ending at 19:20 UTC on June 16, 2026. The impact was split into two primary phases:

  • Phase 1 (17:23 UTC to 18:00 UTC): All Sonnet and Opus models were affected, with error rates reaching approximately 10%.
  • Phase 2 (18:00 UTC to 19:20 UTC): The issue persisted specifically for Opus 4.8, which maintained an average error rate of 10%.

Affected Services and Models

The disruption was widespread, impacting multiple access points and specific model versions. The following services were affected:

  • User Interfaces: claude.ai
  • Developer Tools: Claude API (api.anthropic.com), Claude Code, and Claude Cowork
  • Specific Models: Claude Sonnet, Claude Opus 4.8, and Claude Haiku 4.5

Resolution Process

Anthropic identified the issue and implemented a fix during the incident window. The resolution process involved an initial fix implementation at 18:00 UTC, followed by a period of monitoring and further adjustments to address remaining errors on Opus 4.8 and Haiku 4.5. The incident was officially marked as resolved at 19:32 UTC after monitoring confirmed the stability of the services.

Sources