DeepYardDeepYard
N

NVIDIA Nemotron-3 Super

120B parameter MoE model optimized for agentic reasoning and tool use workflows

commercialFree

About

NVIDIA's flagship 120B parameter language model with latent Mixture-of-Experts architecture specifically optimized for multi-step reasoning and agentic workflows. Trained on diverse datasets including instruction following, conversational tool use, and competitive programming, making it particularly suited for complex agent tasks requiring planning and tool orchestration. Available through HuggingFace with commercial licensing.

Details

Typecoding-agent
Deploymentself-hosted
Supported Models

Tags

autonomoustool-useopen-sourcemulti-agentorchestration