403 "Authorization failed" on integrate.api.nvidia.com/v1/chat/completions while /v1/models returns 200 — request to enable Public API Endpoints on

AI & Data Science NVIDIA NIM Access/Accounts
,
1

Hi,

My newly created nvapi- key authenticates correctly against the model listing
endpoint but is rejected on every inference call. This looks like the missing
“Public API Endpoints” permission on my personal organization, which I’ve seen
reported by several other users in this category.

Working — GET /v1/models returns HTTP 200 and lists 82 models:

curl -s -o /dev/null -w ‘%{http_code}\n’
https://integrate.api.nvidia.com/v1/models
-H “Authorization: Bearer $NVIDIA_NIM_API_KEY”

200

Failing — POST /v1/chat/completions returns HTTP 403 on every model:

curl -s https://integrate.api.nvidia.com/v1/chat/completions
-H “Authorization: Bearer $NVIDIA_NIM_API_KEY”
-H “Content-Type: application/json”
-d ‘{“model”:“nvidia/nemotron-3-super-120b-a12b”,“messages”:[{“role”:“user”,“content”:“hi”}],“max_tokens”:5}’

{“status”:403,“title”:“Forbidden”,“detail”:“Authorization failed”}

I tested four models to rule out a model-specific issue:

Model Result
nvidia/nemotron-3-super-120b-a12b 403 Authorization failed
moonshotai/kimi-k3 403 Authorization failed
moonshotai/kimi-k2.6 403 Authorization failed
deepseek-ai/deepseek-v4-pro 410 Gone (model EOL 2026-08-07)

The 410 on the EOL model confirms the request reaches the routing layer, so the
403 is an authorization decision and not a bad endpoint or malformed request.

Details:

  • Key prefix: nvapi-kic (70 characters), created 2026-09-13
  • Account e-mail: gmoraes.media@gmail.com
  • Org: personal / default (no organization switcher shown in my account)
  • Timestamp of the failing call: 2026-09-13 07:05:52 UTC
  • Response header: content-type: application/problem+json

Could you please enable the “Public API Endpoints” permission for my personal
organization? Happy to provide the full key or any account identifier through a
private message.

Thanks.

2

Hello NVIDIA team,

I am experiencing the same issue described in this topic.

My API key successfully works with GET /v1/models and returns the model list, but every request to POST /v1/chat/completions returns:

403 Forbidden — Authorization failed

I am using a personal NVIDIA Developer account and an active API key for development/testing.

Could you please check and enable Public API Endpoints access for my personal organization?

I have not included my API key for security reasons.

Thank you.