Anthropic's flagship generative AI, Claude, suffered a roughly 45-minute service outage on June 18, 2026, before being restored. Throughout the month, incidents of elevated error rates centered on Opus 4.8 have recurred.
June 18, 2026 · Anthropic Claude
Claude's "Success Tax": 45-Minute Outage Caps a Turbulent June
A surge in demand has outpaced infrastructure. claude.ai, the API, Claude Code and Cowork went dark for ~45 minutes — the latest in a string of incidents that has developers questioning long-term reliability.
~45min
Duration of the June 18 outage (06:55–07:40 UTC)
10+
Significant disruptions reported over 12 days
~10%
Error rates seen across Sonnet & Opus during outages
Recovery times vary by model — June 5 incident
Same outage, very different time to restore. Each block ≈ 12 minutes.
The flagship Opus 4.8 took roughly 6.5× longer to recover than Opus 4.6.
90-day uptime by service
Bars scaled to the full 0–100% range.
Developer frustration
"Opus 4.8 went down and stopped my work"
API-dependent workflows halted, productivity dropped
Refactoring & agentic tasks hit hardest; some fell back to manual coding
Pragmatic takes
Switched to Sonnet to cope with 529 Overloaded errors
"Stick to mature models in production, avoid new ones"
Some credit Claude's past stability record
The core tension
Demand growth is outrunning infrastructure expansion — the "success tax."
Heavy developer & enterprise reliance — API, Claude Code, Claude for Work — amplifies the ripple.
Repeated incidents raise reliability concerns and even talk of a class-action.
Rivals like ChatGPT & Gemini face outages too — scaling is an industry-wide challenge.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…