NVIDIANemotron 3
Nemotron 3 Super 120B-A12B
Nemotron 3 Super is a text reasoning model with 120B total and 12B active parameters.
What the release provides
It uses hybrid LatentMoE layers, Mamba-2, attention and multi-token prediction. The card lists context up to one million tokens. This text release is separate from the later 3.5 Super VL model.Nemotron 3 Super 120B-A12B official model card
How to use it well
Practical guidance. For ticket handling, provide the current ticket, relevant history and the allowed actions. Keep tool permissions outside the model. Review whether a smaller input and a clearer action contract solve errors before increasing model size.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16 |
| Release license | NVIDIA Nemotron Open Model LicenseNemotron 3 Super 120B-A12B official model card |