nvidia models fine (original) (raw)

I can pull and run the gpt-oss-120b NIM container, but the model manifest download fails with HTTP 402 on the nim/team/openai artifact path. The same NGC Personal Key successfully downloads and serves nim/nvidia models on the same host, so the key itself works — the gap seems specific to the openai team’s model resources for oss-120b and not oss-20b as of v2.0.5

What I’ve confirmed:

list-model-profiles runs fine and shows the runnable profile 6d8714cc… (vllm-mxfp4-tp1-pp1-72.0).
nvcr.io/nim/nvidia/nemotron-3-super-120b-a12b:2.0.5 downloads and serves normally with this exact key.
Only nvcr.io/nim/openai/gpt-oss-120b:2.0.5 402s, on the model files path.
Active entitlement: NVIDIA Developer Program. NGC catalog page shows gpt-oss-120b tagged for both Developer Program and NVIDIA AI Enterprise.

Environment:

NIM version 2.0.5
GPU: DGX Spark, single node
Key type: NGC Personal API Key, with NGC Catalog in services

Question: What’s required to download the nim/openai team model artifacts — a specific subscription (NVAIE?), a team-sharing step, or a role assignment (e.g. “Public API Endpoints User”)? I can see that the gpt-oss-120b model is available for pull under the Developer Program, or does it require an additional entitlements?

hmani May 31, 2026, 5:48am 2

Any update on this please? as we are also getting the same error?

@hmani I havent seen any change on my end. I am still receiving the error.

hmani June 1, 2026, 1:31pm 4

if have the model in the cache its using the same. But if you download the models from the nvidia its giving error.

XL42 June 3, 2026, 10:27pm 5

we are getting the same issue as well. was fine a few months ago. was able to get model downloaded but got this 402 when downloading cache.

They just dropped 2.0.6 today so will test with that today.

XL42 June 4, 2026, 7:54pm 7

tried and still the same

I am still getting the same error as well. Does anyone at NVIDIA respond?