The majority of linear layers use NVFP4 for weights, activations, and gradients
8 more from this model
We apply post-training quantization (PTQ) using Model-Optimizer to quantize the Nemotron 3 Ultra checkpoint to NVFP4
open-sourcing the training recipes, data, and RL environments. Checkpoints ... NVFP4 quantized
use NVIDIA’s ultraefficient 4-bit NVFP4 training format on the NVIDIA Blackwell architecture
We release a single NVFP4 checkpoint for Nemotron 3 Ultra
We release a single NVFP4 checkpoint for Nemotron 3 Ultra
These figures use the final NVFP4 weights that NVIDIA recommends for inference
NVFP4 quantized checkpoints
Pretrained in NVFP4.