Claude Status – Elevated errors for multiple models

If you use Anthropic's Claude API, you experienced service disruptions and increased error rates, potentially impacting your applications. This incident highlights the ongoing challenge of maintaining high availability for complex AI models and the potential for cascading failures across different model versions.

On September 22, 2026, Anthropic experienced an outage affecting multiple Claude models, including Claude Fable, Claude Mythos, and Claude Opus. The incident, which began with elevated errors reported around 00:57 UTC, impacted services such as claude.ai, the Claude API, Claude Code, and Claude Cowork. Engineers identified the cause of the errors by 01:17 UTC and worked to implement a fix. By 01:35 UTC, requests to Claude Fable 5, 5.1, and Mythos 5, 5.1 had returned to normal success rates. The company continued to work on resolving remaining errors affecting Claude Opus 5, and reported normal success rates across all affected models by 02:11 UTC, stating they were monitoring the situation closely.

This incident underscores the inherent fragility and complexity of large-scale AI model deployments, even from established providers like Anthropic. The wide-ranging impact across multiple model families and services, including the core API and web interface, indicates a potentially systemic issue rather than an isolated bug. The fact that specific models like Opus 5 experienced lingering problems suggests that the underlying cause may require significant architectural adjustments to fully remediate, posing a challenge for rapid and complete resolution.

The elevated error rates and subsequent recovery timeline, spanning several hours, can disrupt workflows for developers and businesses relying on Claude for real-time applications or critical tasks. Users may have encountered unresponsiveness or incorrect outputs, forcing them to implement fallback mechanisms or delay operations. This event serves as a reminder of the critical need for robust monitoring, rapid incident response, and clear communication from AI providers to mitigate the downstream effects on their customers' services and reputations.