Claude outage hits multiple models again

Anthropic’s Claude AI is showing signs of overload again.

On Monday morning, users trying to work with Claude encountered failed requests and a “529 Overloaded” error. The message says the problem is a temporary server-side issue, then sends users to Anthropic’s status page. A screenshot shared with TNW shows the error appearing in the Claude interface.

Anthropic’s own incident report is broader, and more cautious. At 05:06 UTC on 24 August, the company said it was investigating elevated errors affecting requests to Claude Mythos 5, Claude Fable 5, Claude Opus 5, and Claude Opus 4.8. The incident also affects claude.ai, the Claude API, Claude Code, and Claude Cowork.

At 05:27 UTC, Anthropic said it had identified the cause and was working on a fix. Its latest update, posted at 06:42 UTC, says the company is still working to resolve the issue. The status page does not yet provide a recovery time or a technical explanation.

The distinction matters. “529 Overloaded” describes what at least some users are seeing, but Anthropic has not said that overload is the root cause of the wider incident. For now, the public record establishes elevated errors across several models and products, not why they occurred.

Claude’s busy August

This is not the first Claude service problem this month. Anthropic’s status history includes a degraded-performance incident affecting multiple models on 18 August, followed by another incident involving Opus 5 and Haiku 4.5 on 19 August. A separate disruption affecting multiple models was reported on 20 August.

Independent coverage has also documented the latest interruption. Frandroid reported that some users could not access Anthropic’s models and that the company had identified the problem. Online discussions on Hacker News and other forums show users posting the same 529 error while trying to use Claude.

Those reports do not establish the scale of the outage. They do show how quickly an infrastructure issue becomes visible: in a broken chat request, a failed coding session, or an API call that needs to be retried.

The cost of shared infrastructure

Claude is no longer one product. It is a family of models exposed through consumer chat, developer tools, an API, and Cowork. When several surfaces fail together, users do not experience separate incidents. They experience one unavailable dependency.

That is the more important story here. Anthropic has identified the cause, but it has not yet explained whether the problem sits in a shared service, model routing, capacity, or another part of the stack. Any of those explanations would be speculation at this stage.

The fix is still being monitored. Until Anthropic publishes more detail, the question is not simply when Claude comes back. It is whether users can build on it without treating failure as part of the interface.

For developers, that distinction may matter later. During the outage, the practical question is simpler: whether to wait, retry, switch providers, or close the laptop and play with the cat.