Grok experienced an abnormal failure, and some users encountered large garbled replies. xAI’s chat robot Grok recently experienced an abnormal failure, and some users were using Grok

Grok experienced an abnormal failure, and some users encountered large garbled replies. Grok, a chatbot owned by xAI, recently experienced an abnormal failure. Some users encountered garbled and meaningless text that continued to be generated when using Grok.com. Affected users said that after asking Grok to generate PDF, the model continuously output multiple paragraphs of incomprehensible content, and some responses also included source links pointing to reinforcement learning research websites. According to reports, affected users mainly use Grok Lite, and the problem was first discovered on Wednesday morning. Testing failed to reproduce, suggesting the glitch may only have affected a small number of users. Some users report that refreshing or reopening the session can usually return to normal, but some people continue to receive garbled characters after refreshing multiple times. The issue currently appears to only occur with direct queries to Grok.com, Grok on X.com is not affected. The official Grok account later confirmed that this was a "rare temporary generation failure" and said that restarting the chat or regenerating the conversation usually returns to normal immediately. The official status page shows that Grok's services are running normally and no related service incidents have been recorded. The failure occurred at a time when xAI has been facing high personnel turnover recently. The Information reported in May that xAI had lost most of its founding team members and at least 50 researchers and engineers in recent months. In July, xAI released the latest basic model Grok4.5, saying it reached "Opus level" and had faster speed, higher token efficiency and lower cost. This anomaly also highlights once again that the stability of large model services in the inference, generation and retrieval links is still an important challenge for large-scale deployment. via AI News (author: AI Base)