Breaking

Anthropic Clears Elevated Error Rates Across Multiple Claude Models in About 90 Minutes

June 23, 2026 at 12:42 EDT

Foundation Models
Infra & Chips

On June 23, 2026, Anthropic said it had resolved elevated error rates affecting multiple Claude models. The disruption lasted roughly one hour and 25 minutes from around 14:08 UTC and affected a broad range of services including claude.ai, the Claude API, Claude Code and Claude Cowork, while Claude for Government, which runs on isolated infrastructure, was unaffected.

June 23, 2026 · Anthropic

An 85-Minute Claude Outage — And a Recurring Pattern of Strain

Elevated error rates hit Claude Opus 4.8 and other models across the full stack — chat, API, Claude Code, Console and Cowork — before success rates broadly returned to normal. It is the latest in a string of 2026 disruptions tied to surging inference load.

85 min

Duration of the outage (14:08–15:33 UTC)

~10%

Peak error rate in the similar June 16 incident

Disruptions in 2026: Mar · Jun 5 · 16 · 22 · 23

90-Day Uptime — small but frequent gaps

Higher columns = higher uptime. Differences are tiny but each fraction of a percent is hours of disruption.

99.39%

API

99.26%

Claude Code

99.1%

claude.ai

The model lineup affected

Claude Opus 4.8

Flagship · top performance

$5 / $25

Claude Sonnet 4.6

Balanced

$3 / $15

Claude Haiku 4.5

Low latency · low cost

$1 / $5

Input / output, per million tokens.

How the incident was handled

Investigating → Identified → Fix applied → Monitoring → Resolved

The friction

"API Error: 529 Overloaded" interrupts coding sessions and agentic workflows. Long Claude Code runs and integrations with weak retry logic can cascade into broader app failures.

The perspective

Outages recover quickly and aren't catastrophic — but reliability is the key to competitiveness. Developers urge fallbacks and exponential backoff to stay resilient.

Serving frontier models means a trade-off that prioritizes performance over stability — a strain shared by OpenAI's GPT and Google's Gemini. How far infrastructure fixes can curb recurring outages is now a focal point for adoption.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.