An Empirical Characterization of Outages and Incidents in Public Services for Large Language Models

Xiaoyu Chu, Sacheendra Talluri, Qingxian Lu +1

People and businesses increasingly rely on public LLM services, such as ChatGPT, DALLE, and Claude. Understanding their outages, and particularly measuring their failure-recovery p…