NVIDIA's new Nemotron model takes the top US open-weight slot, Artificial Analysis says
The benchmark group says the 550B-parameter model scores 47.7 on its Intelligence Index, with weights posted on Hugging Face.
By Ryan Merket · Published
Why it matters
NVIDIA is using open weights to compete higher up the AI stack: not just selling GPUs, but offering models that can influence where developers build and deploy inference workloads.
NVIDIA has released Nemotron 3 Ultra, a new open-weight model that Artificial Analysis (@ArtificialAnlys) says is now the strongest US open-weight model on its benchmark, according to a thread on X published Thursday. NVIDIA also pointed to the model in an official X post and its GTC Taipei/Computex 2026 news post.
Artificial Analysis said Nemotron 3 Ultra scored 47.7 on its Artificial Analysis Intelligence Index, ahead of what it described as the next strongest US open-weight models, including Gemma 4 31B. The benchmarking group also said the model has "leading speed for its intelligence," a claim tied to its own comparison framework rather than an independently audited standard.
The model listing on Artificial Analysis identifies it as NVIDIA Nemotron 3 Ultra 550B A55B. NVIDIA has also posted the model weights on Hugging Face, giving developers a path to inspect and run the BF16 release rather than relying only on hosted access.
The timing matters because US open-weight model leadership has become a strategic contest, not just a benchmark table. NVIDIA already sells the hardware most AI labs use; with Nemotron 3 Ultra, it is also pushing further into the model layer where performance, inference speed and deployment control are increasingly part of the same buying decision.