Claude sees elevated errors across services | as Kimi K3, Qwen3.8-Max, Qwen-Image-3.0 ship back-to-back
TL;DR
Anthropic's Status Page on July 21 reported Claude services experiencing elevated errors and degraded performance, incident v35b71rthzhp. Users reported chat latency, timeouts, and 429 rate limits. The incident coincides with back-to-back releases of Kimi K3, Qwen3.8-Max and Qwen-Image-3.0 from Chinese labs.
Anthropic's official Status Page on July 21 reported Claude services experiencing elevated errors and degraded performance. Incident ID v35b71rthzhp; scope and specific affected endpoints not detailed in the initial notice.
It's Claude's Nth infrastructure blip this stretch — the last widespread outage was mid-June's Claude Code 429 misfire storm. Claude 3.7 Sonnet and Claude Opus 4.7 users reported chat latency, timeouts, and requests blocked by rate limits.
Timing is notable — Kimi K3 just open-sourced, Qwen3.8-Max, Qwen-Image-3.0 three Chinese frontier releases in a row, while Anthropic is running elevated errors. API customers wire Claude into production, and even a few minutes of degradation triggers downstream fallback — OpenRouter, LiteLLM and similar routing layers auto-shift traffic to backup models.
Anthropic hasn't published a root cause or ETA. From past patterns, elevated errors like this usually stem from upstream dependencies (AWS Bedrock, GCP Vertex) or internal inference-cluster scheduling issues, not the model weights themselves.
Claude Code users feel it first — the CLI is uniquely latency-sensitive, and one stutter gets flagged and complained about faster than any API SLA breach.
via Claude Status
It's Claude's Nth infrastructure blip this stretch — the last widespread outage was mid-June's Claude Code 429 misfire storm. Claude 3.7 Sonnet and Claude Opus 4.7 users reported chat latency, timeouts, and requests blocked by rate limits.
Timing is notable — Kimi K3 just open-sourced, Qwen3.8-Max, Qwen-Image-3.0 three Chinese frontier releases in a row, while Anthropic is running elevated errors. API customers wire Claude into production, and even a few minutes of degradation triggers downstream fallback — OpenRouter, LiteLLM and similar routing layers auto-shift traffic to backup models.
Anthropic hasn't published a root cause or ETA. From past patterns, elevated errors like this usually stem from upstream dependencies (AWS Bedrock, GCP Vertex) or internal inference-cluster scheduling issues, not the model weights themselves.
Claude Code users feel it first — the CLI is uniquely latency-sensitive, and one stutter gets flagged and complained about faster than any API SLA breach.
via Claude Status
