Hello NVIDIA team,
I am experiencing a 403 “Authorization failed” error when attempting
to call models via https://integrate.api.nvidia.com/v1. The API key
is valid and I can successfully list available models, but any
inference request is rejected.
Account details:
- Email: the same as my forum account
- Organization: Personal
- Endpoint tested: https://integrate.api.nvidia.com/v1/chat/completions
- Model tested: deepseek-ai/deepseek-v4-flash
- Client: Hermes Agent (OpenAI-compatible client)
Error response:
HTTP 403 Forbidden
{“status”:403,“title”:“Forbidden”,“detail”:“Authorization failed”}
Steps already taken:
- Generated a fresh API key from Try NVIDIA NIM APIs
- Verified the key format is correct (nvapi- prefix) and passed
correctly via Authorization: Bearer header - Confirmed model listing works, but chat completion calls fail
When creating a new API key, I only see the “NGC Catalog” permission
option available — “Public API Endpoints” does not appear in the
permission list for my Personal organization.
Could you please enable the “Public API Endpoints” permission for
my account so I can access the hosted serverless NIM inference API?
Thank you.