ModelNemotron-Ultra
Nemotron-3-Ultra-550B-A55B
An ultra-large-scale foundation model in NVIDIA's Nemotron open model family, employing a Mixture-of-Experts architecture with 550 billion total parameters and 55 billion activated parameters. Its weights, training data, and training recipe are all publicly released.
Timeline (last 90 days)
Sep 29
The competition programming model is fine-tuned from Nemotron-3-Ultra-550B-A55B and uses the NVFP4 low-precision quantization format.
Unverified50%