NVIDIA Developer Program member here. Every POST request to https://integrate.api.nvidia.com/v1/chat/completions returns HTTP 403 with the body {“status”:403,“title”:“Forbidden”,“detail”:“Authorization failed”}.
What I have already verified:
(a) Two API keys generated from Try NVIDIA NIM APIs, both ACTIVE and not expired.
(b) One key was freshly generated and never used, so rate limiting is not possible.
(c) GET /v1/models with the same key returns HTTP 200, so authentication itself passes. Only POST /v1/chat/completions returns 403.
(d) Tried several models that are all listed as “Free Endpoint: Available” on build.nvidia.com, including openai/gpt-oss-20b, minimaxai/minimax-m3 and GLM-5.2. All return the same 403.
(e) The response is 403 Forbidden, not 401 Unauthorized and not 429 rate limited.
(f) My account page shows “Your API Rate Limit: Up to 40 rpm”.
(g) I only have a single personal organization, so there is no other organization to switch to.
(h) Same result from curl and from an OpenAI-compatible client.
This matches the known issue where the “Public API Endpoints” permission is missing on a personal organization, which has been reported many times in this category.
Could you please enable the Public API Endpoints permission for my organization?
Account email: m.fatihyerlikaya@gmail.com
Organization: Fatih Yerlikaya (personal)
Thank you.