Claude Outage Hits Multiple Models as Anthropic Reports Elevated Error Rates
5 min. read 
Published on
Anthropic reported elevated error rates across multiple Claude models on Tuesday, June 23, 2026, affecting users who rely on Claude for chat, coding, API work, and developer workflows.
The incident appeared on the official Claude Status page as “Elevated error rate across multiple models.” Anthropic first said it was investigating the issue at 14:19 UTC, then said it had identified the cause and was implementing a fix at 14:25 UTC.
Access content across the globe at the highest speed rate.
70% of our readers choose Private Internet Access
70% of our readers choose ExpressVPN
Browse the web from multiple devices with industry-standard security protocols.
Faster dedicated servers for specific actions (currently at summer discounts)
By 14:53 UTC, Anthropic said a fix had been implemented and that its team was monitoring the results. A later update at 15:28 UTC said the company continued to monitor for remaining issues, suggesting services were recovering but not yet fully cleared at that point.
Claude services saw elevated errors, not a confirmed total shutdown
The status record described the incident as elevated error rates, which usually means some requests fail while others continue to work. That is different from a full outage where the entire service becomes unavailable for all users.
For users, the symptoms can still feel like an outage. Claude may fail to respond, return server errors, stall during generation, or break inside tools such as Claude Code. Developers calling the API may also see higher failure rates during affected windows.
Anthropic’s Claude API error documentation explains that 500 errors indicate internal system issues, 504 errors mean a request timed out, and 529 errors mean the API is temporarily overloaded.
| Incident detail | What Anthropic reported |
|---|---|
| Incident title | Elevated error rate across multiple models |
| Date | June 23, 2026 |
| First acknowledged | 14:19 UTC |
| Cause identified | 14:25 UTC |
| Fix implemented | 14:53 UTC |
| Monitoring update | 15:28 UTC |
The disruption affected users across Claude products
The incident mattered because Claude now supports more than one type of workflow. Users access the models through the web app, the developer console, the API, Claude Code, and newer work products.
During a multi-model incident, users may see errors even if they switch prompts, restart a browser, or retry the same request. If the problem sits in shared model infrastructure, normal troubleshooting on the user’s device may not fix it.
Third-party status monitors such as StatusGator’s Claude tracker also showed Claude-related disruption on June 23, including recent incidents affecting Claude Opus 4.8 and broader model availability.
June has been a busy month for Claude reliability incidents
Tuesday’s incident followed several other Claude reliability problems in June. The official Claude incident history lists repeated elevated-error events across June 2026, including several incidents tied to Claude Opus 4.8.
Earlier on June 23, Anthropic resolved a separate incident titled “Elevated errors for Claude Opus 4.8,” which ran from 06:28 UTC to 08:45 UTC. On June 22, the company also logged elevated-error incidents affecting Opus 4.8 and other Claude models.
One of the more detailed recent entries came on June 16. The Claude status history said all Sonnet and Opus models reached an error rate of around 10% during part of the incident, before Opus 4.8 alone continued at an average error rate of about 10% until recovery.
What users should do when Claude is down or unstable
Users should first check Anthropic’s status page before spending time troubleshooting local settings. If an active incident appears, the issue likely sits with the provider rather than the user’s browser, app, API key, or network.
Developers should avoid aggressive retry loops during outage windows. Repeated requests without delay can add load to a strained service and may make failures worse for the user’s own application.
The Anthropic API error guide says 529 errors can occur when APIs experience high traffic across all users. For production systems, that makes backoff, retry limits, and model fallback planning important.
- Check the official Claude status page before debugging local settings.
- Retry failed requests with delay instead of sending rapid repeated requests.
- Switch models if only one Claude model has elevated errors.
- Pause time-sensitive Claude Code work if the API is unstable.
- Use a fallback AI provider for critical production workflows.
- Monitor API error codes separately from user-facing Claude web issues.
Why Claude outages matter for businesses and developers
Claude has become part of daily work for software teams, support teams, writers, analysts, and enterprises. When error rates rise, the impact can spread beyond casual chatbot use and into production pipelines, customer service systems, code review, and automation.
This is why status visibility matters. A partial outage can create confusing symptoms because some users may still get responses while others see repeated failures. Teams need a clear way to tell whether the problem comes from their own system or the model provider.
Monitoring services such as StatusGator can help teams track incidents across multiple AI services, but Anthropic’s own Claude Status page remains the most important source for official incident updates.
FAQ
Anthropic reported elevated error rates across multiple Claude models on June 23, 2026. The incident was best described as degraded availability or partial disruption, not a confirmed full shutdown of every Claude service.
Anthropic said it had identified the cause and was implementing a fix, but the public status update did not provide a detailed technical root cause at the time of the incident updates.
The incident affected multiple Claude models and could impact users of Claude web, Claude Console, Claude API, Claude Code, and related products depending on the request path and model used.
An elevated error rate means more requests than usual are failing. Some users may still get responses, while others may see failed generations, server errors, timeouts, or overloaded API messages.
Developers should check the official status page, avoid rapid retry loops, use exponential backoff, monitor API error codes, and route critical workloads to fallback models or providers when needed.
Read our disclosure page to find out how can you help VPNCentral sustain the editorial team Read more
Improve this guide
User forum
0 messages