Anthropic's Claude models threw elevated errors twice on 25 October, once for 1h 39m affecting several models and once for 33m affecting Claude Fable 5.1. Both faults sat on Anthropic's side and were fixed before you could have done anything about them.
Started
06:14 UTC
Duration
Lasted 1h 39m
Source
IsDown
Next time
Anthropic is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
Affected components: claude.ai, api.anthropic.com
What happened?
The first incident started at 06:14 UTC when Anthropic began investigating elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5. Thirteen minutes later Anthropic said it had identified the cause, narrowing the affected model to Claude Sonnet 5.5, and said a fix was coming. A fix was deployed at 06:35 UTC, and the incident was marked resolved at 07:53 UTC, with Anthropic stating the actual impact window ran from 06:05 UTC to 06:35 UTC, shorter than the time the status page stayed open. That gap between the stated impact window and the resolution timestamp is on the record as given, not something we're guessing at. The second incident started at 08:17 UTC, this time narrower in scope, naming only Claude Fable 5.1. Anthropic's update sequence moved fast: investigating and identified were logged in the same minute, a fix was deployed by 08:30 UTC, and the incident was resolved at 08:49 UTC, with the stated impact window running 08:08 to 08:29 UTC. Both incidents affected claude.ai and api.anthropic.com, the only two components listed for the day. No user reports were filed during either window. Anthropic's status page is the only source for both incidents, and in both cases it names the affected models and timing but gives no explanation for what caused the errors underneath.
Learning
Anthropic's own status page is the account of record here, and a status page is written by the provider about the provider's own infrastructure, which means it can label and time an incident precisely while telling you nothing about whether the fault ever reached your product. Elevated error rates on specific models, like Claude Sonnet 5.5 or Claude Fable 5.1, happen at the model-serving layer, somewhere between your API calls and Anthropic's own backend, a layer your own uptime checks on your product cannot see at all. If your monitoring only watches your own endpoints, a run of failed completions during one of these windows looks like a mystery on your end when it was actually logged, fixed and resolved on Anthropic's side within the hour. Over the past 90 days Anthropic has logged 52 such incidents with a median duration of 56 minutes, so this kind of short, model-specific error spike is a recurring pattern rather than a one-off. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.