Claude reports elevated error rates across Sonnet, Opus, and Haiku models
Anthropic's incident report documents two phases of elevated errors on June 16, 2026, affecting multiple Claude models and services.
1 source · cross-referenced
- Elevated error rates affected Claude Sonnet, Opus, and Haiku models on June 16, 2026.
- Two phases of disruption were reported: one affecting all Sonnet and Opus models, and a second focused on Opus 4.8.
- The incident impacted claude.ai, the Claude API, Claude Code, and Claude Coworker.
Anthropic reported an incident on June 16, 2026, involving elevated error rates across multiple Claude models. The first phase lasted from 10:23 PT / 17:23 UTC to 11:00 PT / 18:00 UTC and affected all Sonnet and Opus models, reaching an error rate of around 10%.
A second phase followed from 11:00 PT / 18:00 UTC to 12:20 PT / 19:20 UTC, during which Opus 4.8 experienced an average error rate of 10%.
The incident impacted core services including claude.ai, the Claude API (api.anthropic.com), Claude Code, and Claude Coworker.
Anthropic confirmed the issue was resolved after implementing a fix and monitoring results. Updates were provided at 17:29 PT, 18:00 PT, 18:15 PT, 19:08 PT, 19:22 PT, and 19:32 PT.
- Jun 16, 2026 · AWS — Machine Learning Blog
AWS SageMaker AI adds container caching to cut generative AI inference scale-out latency by up to half
Trust79 - May 21, 2026 · TechCrunch
Spotify launches ElevenLabs-powered audiobook creation tool for independent authors
Trust54 - May 20, 2026 · Hugging Face
Hugging Face releases six Ettin reranker models with distillation training recipe
Trust74