Nvidia Releases Nemotron 3 Super, a 120B Open AI Model Built for Agentic Workloads

Nvidia released Nemotron 3 Super, a 120-billion-parameter open Mixture-of-Experts (MoE) model that activates ~12.7B parameters per forward pass and supports native 1‑million‑token context windows. Nvidia claims up to 7.5x throughput versus Qwen3.5-122B-A10B and material gains versus prior Nemotron Super and GPT-OSS-120B, with optimized NVFP4 quantization for Blackwell GPUs and strong inference performance on B200 hardware. The open licensing and availability of checkpoints, training data and inference paths (cloud and on‑prem) broaden adoption prospects for AI developers and could lift demand for Nvidia compute (and related ecosystem players), signaling a potentially bullish market impact for NVDA.OQ and GPU-driven AI infrastructure providers.