Hello NVIDIA Support Team,
I am experiencing inference failures with nvidia/nemotron-3-ultra-550b-a55b through both the NVIDIA Build Playground and the direct NVIDIA API.
Diagnostic results:
-
NVIDIA API TCP 443: Successful
-
GET
/v1/models: HTTP 200 in 0.53 seconds -
Target model listed: Yes
-
POST
/v1/chat/completions: Timeout after 45 seconds without an HTTP response -
NVIDIA Build Playground:
Retries exhausted: 3/3 -
Hermes Agent integration: Request interrupted while awaiting API completion
Could you investigate whether this model’s hosted inference endpoint is experiencing availability issues, or whether my account has inference-specific restrictions?
Please also advise whether Public API Endpoints permissions need to be checked.
Thank you.