Nemotron 3 Super
NVIDIA Nemotron 3 Super is a 120B-parameter hybrid Mixture-of-Experts (MoE) model designed for compute efficiency and accuracy in complex multi-agent applications. It features a hybrid Mamba-Transformer architecture with multi-token prediction (MTP) and a 1M token context window for long-term coherence, cross-document reasoning, and multi-step task planning. The model leverages latent MoE to activate only 12B parameters during inference, enabling high intelligence and generalization at reduced computational cost. It is trained across 10+ environments using multi-environment reinforcement learning and achieves leading accuracy on benchmarks such as AIME 2025, TerminalBench, and SWE-Bench Verified.
What is Nemotron 3 Super?
Nemotron 3 Super is an AI model from NVIDIA that Agent Mag tracks for pricing, context window, modalities, benchmarks, and API compatibility. Builders can use this page to compare Nemotron 3 Super against other models for agent workflows and production deployments.
NVIDIA Nemotron 3 Super is a 120B-parameter hybrid Mixture-of-Experts (MoE) model designed for compute efficiency and accuracy in complex multi-agent applications. It features a hybrid Mamba-Transformer architecture with multi-token prediction (MTP) and a 1M token context window for long-term coherence, cross-document reasoning, and multi-step task planning. The model leverages latent MoE to activate only 12B parameters during inference, enabling high intelligence and generalization at reduced computational cost. It is trained across 10+ environments using multi-environment reinforcement learning and achieves leading accuracy on benchmarks such as AIME 2025, TerminalBench, and SWE-Bench Verified.