I have a Switchyard setup connected to a Bedrock Mantle endpoint for GPT 5.6 Luna. Since earlier today, even very simple requests such as "print hello" have been taking 30–60 seconds or failing to return at all. I'm testing with direct curl requests against the Mantle endpoint. The regular, non-Mantle endpoint is responding normally. Is anyone else seeing similar throttling or latency?
2 Answers
This may be a broader Mantle-side issue rather than a problem with your request or curl setup. If the same basic calls are timing out across multiple attempts while the standard endpoint remains healthy, the endpoint type is the most likely difference.
Our provisioned-capacity workloads have been running at their usual speed, so this doesn’t appear to affect every Bedrock service path. That said, it doesn’t rule out a problem isolated to the Mantle endpoint.
That makes sense, but the provisioned-capacity result may not tell us much about Mantle specifically. The contrast I’m seeing is between Mantle and the regular endpoint.