Anthropic reported an increase in errors across five models

Anthropic launched an investigation into increased error rates across five Claude models on October 25, 2026. The company published an incident record titled “Increased errors across multiple models” on the Claude Status page at 09:14 Turkey time. The record lists the affected models as Claude Mythos 5.1, Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, and Claude Haiku 5.5.

Which services were affected

According to the incident record, the outage covers claude.ai, Claude API (api.anthropic.com), Claude Code, and Claude Cowork. On the Status page, these four components are marked as a partial outage. Claude Console (platform.claude.com) and Claude for Government appear to be operational.

In the record, Anthropic said it was investigating the cause of the errors and would share an update as soon as possible. Seven minutes after the record was published, the incident was still in the “investigating” stage, and no additional statement had been made.

90-day uptime data

Historical data on the Status page shows that outages like this do not significantly distort the overall picture. Over the past 90 days, claude.ai reached 99.58% uptime, Claude API 99.64%, Claude Code 99.58%, and Claude Cowork 99.57%. Claude Console is at 99.84%, while Claude for Government is at 100%.

Recurring waves in a week

This incident doesn’t come in isolation. In the Status history, three separate incidents were resolved on October 24, 2026: the Claude Fable 5 and Fable 5.1 error was closed at 08:53 Turkey time, the Claude Opus 5.5 error at 03:37, and the Fable 5.1, Sonnet 5.5, and Haiku 5.5 error at 03:11. Records show that the impact lasted about half an hour in both incidents that day.

Before that, on October 22, Connectors-related errors affected claude.ai and Claude Cowork; the investigation began at 06:50 Turkey time and the incident was resolved at 11:07. On October 21, a performance drop was reported for Claude Fable 5. A similar “increased errors across multiple models” incident also occurred on August 20, 2026, and was closed in 26 minutes.

What it means for users

For teams running automated workflows that depend on Claude, outages like this are a reminder of how critical error handling and retry logic are. Seeing similar errors across multiple models on the same day makes it sensible to keep a fallback plan for a second model or endpoint instead of relying on a single provider. Following Anthropic’s updates in the incident record is the safest way to understand whether the error is model-related or infrastructure-related.

Source: Unite.AI