Degraded DeepSeek V4.1 Flash on US inference
Affected: inference
Timeline
- Closed operatorThe observation period is complete, and this incident is now closed.
- Resolved operatorAll clear: DeepSeek V4.1 Flash on the US inference endpoint has recovered, and normal service is restored. We are continuing to monitor for any recurrence. Thank you for your patience.
- Reported operatorCustomers using DeepSeek V4.1 Flash through the US inference endpoint may experience elevated errors and slower responses.
- Investigating operatorThe US DeepSeek V4.1 Flash service remains below our normal success threshold. We are investigating and continuing to monitor the service. EU inference is healthy, and no failover is currently active.