A Postmortem of Three Recent Issues

Anthropic's September 2025 postmortem explains how a routing fault gave users different experiences of the same model. Some Sonnet 4 requests reached servers configured for a different context length, degrading responses. Because routing tended to send follow-up requests to the same server, an affected user could encounter repeated problems while another saw normal behavior. A load-balancing change increased exposure. The company corrected the routing and said it would run quality evaluations continuously on production systems. The incident shows why model testing must include the infrastructure that actually handles users' requests.

“We never reduce model quality due to demand, time of day, or server load.”

“The problems our users reported were due to infrastructure bugs alone.”

“A small subset of users that asked a question in English might have seen "สวัสดี" in the middle of the response, for example.”

“Model quality is non-negotiable, so we accepted the minor efficiency impact.”