Recipe ·

Nemotron 3 Super

NVIDIA's Nemotron 3 Super is a 120-billion-parameter open model that uses a hybrid Mamba-Transformer MoE architecture to deliver high throughput for agentic AI systems.

Techniques inside

Median research → prod

Fastest adoption

Slowest adoption

Ingredient list

#1Mamba / Selective State Space Models
Invented by CMU · 2023-12 · Velocity 2y
“Nemotron 3 Super uses a hybrid Mamba-Transformer MoE architecture.”
architecturehigh
#2Transformer Self-Attention
Invented by Google · 2017-06 · Velocity 9y
“Nemotron 3 Super uses a hybrid Mamba-Transformer MoE architecture.”
architecturehigh
#3Mixture of Experts (Sparse MoE for LLMs)
Invented by Google · 2017-01 · Velocity 9y
“Nemotron 3 Super uses a hybrid Mamba-Transformer MoE architecture.”
architecturehigh

This recipe is part of the gentic.news Deployment Atlas. Every ingredient has an origin paper + evidence. Methodology is public. Dataset is CC BY 4.0.