Global AI failure tracking: Claude service back to normal, ChatGPT and Codex had high error rates, Grok maintenance machine

OneMillion AI reported that global AI failure tracking indicated that Claude services were back to normal, but ChatGPT and Codex had a high error rate, and OpenAI had implemented mitigation measures and monitored recovery. There are multiple model errors today, affecting Mythos/Fable 5.1, Mythos/Fable 5 and Opus 5, 4.8, 4.6. Officially, at 00:16 UTC+8, the impact was declared to be over, and the web pages, the console, API, Claude Code and Cowork are all now operational. The Grok official status page shows that the main portals for Grok to the users are still downlined and that services in the event of failure include Grok iOS, Grok Android, Grok Web version, Grok Build, Office/Workspace plugins, Eastern United States API and Western United States API. Normal services include login, European API, API console, documents and XAI official network。