NVIDIA has introduced three groundbreaking models in the Llama-Nemotron series: LN-Nano 8B, LN-Super 49B, and LN-Ultra 253B. Tailored for inference tasks, these models boast high-throughput capabilities, robust reasoning skills, and extended context handling. Notably, the LN-Ultra model, operating on a single 8xH100 node, surpasses DeepSeek-R1 in performance, solidifying its position as one of the most potent open-source models for scientific reasoning on the market today.