Opus 4.6 had elevated errors on April 19, 2026
This incident reinforces the difference between model-specific API instability and a full claude.ai outage. For some users, Claude was unreliable; for others, the problem may have looked invisible.
Published: Apr 19, 2026, 6:44 PM EDT
Updated: Apr 21, 2026, 8:40 AM EDT
Incident started: Apr 19, 2026, 5:29 PM EDT
Crowd spike noticed: Apr 19, 2026, 5:40 PM EDT
Resolved: Apr 19, 2026, 6:44 PM EDT
Affected surface: Opus 4.6 requests
Official updates described elevated errors on Opus 4.6 and later marked the model incident resolved.
- The issue looked concentrated in one model rather than all Claude surfaces simultaneously.
- Model-level instability can create confusing reports because some users continue to see normal behavior elsewhere.
- This type of incident is especially relevant for the Claude API and heavy Claude Code usage.
A reviewed note on the April 19, 2026 Opus 4.6 incident that looked like a model-specific outage rather than a full Claude-wide failure.
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: