Anthropic Resolves Claude Opus 5 Error Surge After Hours-Long Disruption
Service outage affected claude.ai, API, Claude Code and Cowork; users reported hallucinations and erratic behavior during the incident.
Anthropic restored normal service for its flagship Claude Opus 5 model on July 27, 2026, after elevated error rates disrupted access across the company's product line for several hours.
The incident affected claude.ai, Claude API (api.anthropic.com), Claude Code, and Claude Cowork. According to Anthropic's status page, investigation began at 11:27 UTC and the company posted its initial incident report at 12:30 UTC. Errors returned to baseline as of 4:47 PST (11:47 UTC).
During the disruption, users took to Hacker News to report symptoms beyond standard error messages. Commenters described increased hallucinations, off-kilter language output, and the model behaving in ways they characterized as "aggressively proactive." One user wrote that Opus 5 was "giving me many more hallucinations, including things that aren't even in the right territory."
The quality complaints add a second layer to what began as a straightforward availability incident. While Anthropic's status communication focused on error rates, user reports suggest the model's outputs may have been degraded in ways that aren't captured by simple success/failure metrics.
Some Hacker News participants used the incident to voice broader frustration with Opus 5's reliability compared to earlier versions, claiming the model is more prone to coding regressions. These comparisons reflect user opinion rather than independently verified performance data.
Service disruptions in flagship models carry outsized impact for enterprises and developers running production workflows on Claude's API. The incident underscores the operational risk that comes with integrating frontier AI models into critical infrastructure, where both availability and output quality matter.
Anthropic did not disclose the root cause of the elevated errors or the reported behavioral anomalies. The company's status page provided minimal technical detail beyond confirming the affected services and the resolution timestamp.
The July 27 incident was the second disruption Anthropic logged that day. An earlier incident resolved at 9:03 UTC preceded the Opus 5-specific problems that began roughly two hours later.