Crowd Signal Monitor
← Back to tracker

Sonnet 4.6 had a prolonged error spike on April 8, 2026

This was a longer model-specific event. It matters because developers can experience a very real outage while the broader Claude site still appears partially functional.

Published: Apr 8, 2026, 2:23 AM EDT

Updated: Apr 21, 2026, 8:40 AM EDT

Incident started: Apr 8, 2026, 2:23 AM EDT

Crowd spike noticed: Apr 8, 2026, 3:06 AM EDT

Resolved: Apr 8, 2026, 5:33 AM EDT

majorModel/API3 hours 10 minutes

Affected surface: Claude Sonnet 4.6 requests

Official updates said Sonnet 4.6 was exhibiting errors above baseline for several hours before the model incident was fully resolved.

  • The incident lasted longer than many of the other model-only notes in this archive.
  • It shows how a single unstable model can create real outage conditions for some developers without crashing every Claude surface.
  • The multi-stage official updates suggest the fix path was more involved than the brief model blips elsewhere in the month.

A reviewed note on the April 8, 2026 Sonnet 4.6 incident where one model family stayed above baseline for several hours.

Review the methodology page for how crowd reports are deduplicated and the status lag explainer for how independent reports differ from provider updates.

If you want the broader archive context, go back to the Claude outage history page.

Related notes in the same incident category: