Anthropic experienced elevated error rates across five of its Claude AI model variants—Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 4.6, and Haiku 4.5—in what the company classified as a major incident. The issue has since been resolved.
Anthropicis a prominent AI safety company that develops Claude, a family of large language models used by developers and enterprises for a range of applications.
The company began investigating the error rates, identified the root cause, and implemented a fix. Opus 4.8, Haiku 4.5, and Opus 4.7 recovered sequentially as the fix was deployed. Anthropic subsequently confirmed that all affected services had stabilized and are now being monitored.
The incident is now closed.