nemotron-3-super:cloud
Upstream reference: Nemotron 3 Super NVFP4
Publisher: NVIDIA
NVIDIA documents a hybrid LatentMoE model with Mamba-2, attention and multi-token prediction, designed for reasoning and collaborative agent applications.
Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.
Published upstream specifications
- Total parameters
- 120 billion Source
- Active parameters
- 12 billion Source
- Published context
- Up to 1M tokens, publisher notation Source
- Named checkpoint precision
- NVFP4 Source
- Upstream license metadata
- NVIDIA Nemotron Open Model License Source
Publisher-described uses
- Text reasoning, retrieval-assisted applications and tool-oriented workloads. Source
Reference limitations
- The named NVFP4 checkpoint, NIM container and hosted service have distinct artifact and terms boundaries; no runtime or availability is inferred. Source
A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.
Publisher sources
- Nemotron 3 Super NVFP4 official documentation — NVIDIA. Checked .
Links
Recorded references do not establish downloadable weights, licensing or current availability.
120B
parameters
NVFP4
quantization
nemotron_h_moe
architecture
262144
context
nvidia-nemotron-open-model-license
license
2026-03-11
released
Where nemotron-3-super:cloud comes from
Company / lab
NVIDIA
See every model and family from this maker.
Model family
nemotron-3-super
Other versions and sizes in this line.
completionthinkingtools