Capacity is partially restored and stable after two incidents in 24 hours, but Anthropic says it isn't fully in the clear yet. Updates will continue on the status page.
Changing the effort level setting invalidates the prompt cache because effort level is communicated to the model via the system prompt. A team member confirmed this is basically correct.
Two separate incidents over the past 24 hours caused elevated errors and reduced availability for some users. Multiple network failures cut into serving capacity while traffic was being rerouted.
