Anthropic experienced a service disruption on Tuesday that produced elevated error rates across multiple Claude models, according to the company’s official status page. By mid-afternoon UTC the company said a fix had been deployed and that it was monitoring systems for further problems.

The incident, logged on status.claude.com as “Elevated error rate across multiple models,” was first acknowledged at 14:19 UTC when Anthropic posted that it was investigating the issue. Within minutes, at 14:25 UTC, the company said it had identified the cause and was implementing a fix.

By 14:53 UTC, Anthropic reported that the fix had been applied and that it had moved into a monitoring phase. A further update at 15:28 UTC said the company was continuing to watch for any additional issues, indicating that core service had largely been restored but that engineers were not yet ready to mark the incident fully resolved.

The disruption was reflected in the status of several Anthropic products, which were marked as experiencing partial outages, including the consumer-facing claude.ai interface, the Claude Console at platform.claude.com, the Claude API, Claude Code, and Claude Cowork. Claude for Government remained operational throughout.

While the incident affected more than one model, the public status record characterizes the impact as elevated error rates and partial outages rather than a complete loss of service. In practical terms, that typically means a portion of user requests failed or returned errors while many requests continued to be served, rather than the platform going dark entirely.

Anthropic did not publish a specific error-rate figure for this particular incident in its initial updates, though the company has disclosed such numbers for past events.

Part of a broader pattern this month

Tuesday’s incident was not isolated. The status history shows a notably busy stretch for Anthropic’s reliability team through June 2026, with repeated incidents concentrated on the company’s flagship model, Claude Opus 4.8.

Earlier the same day, a separate incident involving elevated errors for Claude Opus 4.8 ran from roughly 06:28 UTC until it was resolved at 08:45 UTC. The day before, on June 22, the company logged multiple incidents, including one affecting several models — Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 4.6, and Haiku 4.5 — that was resolved in the early hours of the morning UTC.

One of the more clearly quantified recent events occurred on June 16, when Anthropic reported a two-phase incident: all Sonnet and Opus models saw error rates reach roughly 10 percent for part of the window, followed by a period in which Opus 4.8 alone averaged about a 10 percent error rate before recovery.

Follow us on Google News, LinkedIn, and X to Get More Instant Updates.