Claude Simultaneous Outage Hits Opus 4.8, 4.7, 4.6, and Sonnet 4.6 — Four Models, One Failure
Anthropic’s status page logged elevated error rates across Claude Opus 4.8, Opus 4.7, Opus 4.6, and Sonnet 4.6 on June 22 — its entire revenue-generating model tier, affected simultaneously.
The breadth narrows the diagnosis. A per-model defect would take down one checkpoint. Four models across two separate generations going offline together points upstream: a routing, load-balancing, or serving layer shared across the Opus family. Haiku 4.5 was not listed in the incident scope.
What Went Down
All four models span Anthropic’s enterprise pricing stack: Opus 4.8 at $15/$75 per million tokens, Opus 4.7 at $5/$25, Opus 4.6 at standard Opus pricing, and Sonnet 4.6 covering the mid-tier. Anthropic’s status page confirmed it was monitoring recovery. The outage was brief. No figures on affected API traffic volume were disclosed.
Context: Infrastructure at Scale
Anthropic recently consolidated all 220,000 GPUs at SpaceX’s Colossus 1 and is committed to Colossus 2 with GB200 hardware. That scale, combined with multi-region expansion to anchor enterprise SLAs, creates dependency chains that can produce exactly this kind of horizontal blast radius — a single shared component upstream of multiple model endpoints.
The timing is notable. Anthropic is tracking $10.9B in Q2 revenue, its first quarter on a trajectory toward profitability, with enterprise agreements spanning KPMG’s 276,000-user deployment, Claude Code production pipelines across major tech firms, and Managed Agents for customer-facing workloads. All are exposed when the shared serving layer fails.
Key Numbers
- Models affected: Claude Opus 4.8, 4.7, 4.6, Sonnet 4.6
- Outside incident scope: Claude Haiku 4.5
- Anthropic Q2 revenue run-rate: $10.9B
- Colossus 1 GPU count: 220,000
- Enterprise exposure: KPMG 276,000 seats, Claude Code, Managed Agents