403 Authorization Failed on NIM API — Public API Endpoints Permission Not Visible

Title: 403 Authorization Failed on NIM API — Public API Endpoints Permission Not Visible

Hello,

I am trying to use the NVIDIA NIM hosted API through the OpenAI-compatible endpoint:

https://integrate.api.nvidia.com/v1

I created an API key from NVIDIA Build using my personal NVIDIA Developer account.

The API key appears to be valid because I can successfully call:

GET /v1/models

and receive a list of available models.

However, every attempt to call:

POST /v1/chat/completions

returns:

403 Forbidden — Authorization failed

I have tested multiple models, including:

  • google/gemma-4-31b-it

  • openai/gpt-oss-20b

Both return the same 403 authorization error.

For example, I am using the OpenAI-compatible Python client with:

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://integrate.api.nvidia.com/v1",
    api_key=os.environ["NVIDIA_API_KEY"],
)

response = client.chat.completions.create(
    model="openai/gpt-oss-20b",
    messages=[
        {
            "role": "user",
            "content": 'Responde solamente con JSON válido: {"estado":"OK"}',
        }
    ],
    temperature=0,
)

The /v1/models request succeeds, but /v1/chat/completions consistently returns:

403 {'status': 403, 'title': 'Forbidden', 'detail': 'Authorization failed'}

I checked my NVIDIA Build account menu, but I only see:

  • Account name

  • API Keys

  • Interactions

  • Sign out

I do not see any Organization, Permissions, Entitlements, or Public API Endpoints setting.

Could you please check whether Public API Endpoints access is enabled for my personal organization/account, or let me know what I need to do to obtain access?

I am using the API for development and testing of a document-processing application.

I have intentionally not included my API key or any other credentials in this post.

Thank you.

I am experiencing the exact same issue after accidentally deleting a working legacy key. My GET /v1/models request succeeds with a 200 OK, but all newly generated keys (as well as keys generated 1-2 weeks ago) return a 403 Forbidden error on POST /v1/chat/completions.

Like the author above, my personal organization menu doesn’t show any Organization, Permissions, or Public API Endpoints configurations to toggle.

Could an administrator please manually refresh and enable the “Public API Endpoints” permission flag for my account backend?

  • Registered Account Email: aliang.nvda@gmail.com
  • API Key Prefix (first 8 characters only): nvapi-3WnI7Nq6

Thank you!

Hello NVIDIA Support,

I am building a personal AI assistant called HOVER using Python and NVIDIA’s hosted API.

My API key successfully retrieves the model list, but requests to the chat-completion endpoint return the same 403 error for multiple models.

Error: 403 Forbidden — Authorization failed

Endpoint: https://integrate.api.nvidia.com/v1/chat/completions

Could you please check whether my personal account has permission to use hosted inference through the Public API Endpoints?

Thank you.