403 Authorization failed — Public API Endpoints access for my personal account

AI & Data Science NVIDIA NIM Access/Accounts
, ,
1

Hello NVIDIA Team,

I am using the NVIDIA Developer Program and an NVIDIA API key for personal development of my engineering project.

My NVIDIA Build Playground works correctly. I can select a model and send prompts successfully, and the model returns a response.

However, the API key cannot perform inference through the public API.

Working:

GET https://integrate.api.nvidia.com/v1/models
→ HTTP 200 OK

Failing:

POST https://integrate.api.nvidia.com/v1/chat/completions
→ HTTP 403 Forbidden

Response:

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

I reproduced this directly from Windows PowerShell, independently of OpenCode.

I tested multiple models, including:

  • nvidia/llama-3.1-nemotron-70b-instruct

  • moonshotai/kimi-k3

  • deepseek-ai/deepseek-v4.1-flash

All inference requests return the same 403 authorization error.

The NVIDIA Build Playground works normally.

I also completed my NVIDIA Developer Program profile.

Could you please check whether my personal organization/account has the required Public API Endpoints permission or hosted inference entitlement enabled?

If it is not enabled, please enable it or let me know what additional account verification is required.

I will not post my API key publicly.

Hello NVIDIA Build Support Team,

I am using the NVIDIA Developer Program and an NVIDIA API key for personal development of my engineering project, ELI.

I am experiencing an inference access issue with the NVIDIA Public API.

My NVIDIA Build Playground works correctly. I can select models, send prompts, and receive responses.

However, API inference does not work with the same API key.

I have now isolated the issue from OpenCode and from my ELI project by testing the API from a completely separate NVIDIA AI Workbench project.

Test results

1. Model catalog

GET:

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

Result:

HTTP 200 OK

The model catalog is returned successfully.

2. Windows PowerShell inference

POST:

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

Result:

HTTP 403 Forbidden

Response:

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

I reproduced this with multiple currently listed models.

3. NVIDIA AI Workbench isolation test

I created a completely separate NVIDIA AI Workbench test project and configured the same API key as a sensitive environment variable.

From the Workbench container:

GET /v1/models

→ SUCCESS

POST /v1/chat/completions

→ HTTP 404

Response:

{“status”:404,“title”:“Not Found”,“detail”:“Function ‘…’: Not found for account ‘…’”}

I am intentionally omitting the account identifier and API key from this email.

Current situation

  • NVIDIA Developer account: active

  • NVIDIA API key: accepted by /v1/models

  • /v1/models: HTTP 200

  • NVIDIA Build Playground: works successfully

  • Windows API inference: HTTP 403 Authorization failed

  • OpenCode: HTTP 403 Authorization failed

  • Clean NVIDIA AI Workbench container: HTTP 404 Function not found for account

  • The issue is reproducible independently of OpenCode and independently of my ELI project

Could you please check whether my personal NVIDIA Developer account/organization has the required Public API Endpoints / hosted inference entitlement / function access enabled?

If additional account verification, organization configuration, or activation is required, please let me know what I need to complete.

I will not share my API key publicly.

Thank you for your help.

Best regards,

Serhat Erbektaş
Founder / CEO
ELI