NVIDIANemotron 3.5
Nemotron 3.5 Super VL 120B-A12B
Super VL combines a hybrid language model with a C-RADIO vision encoder. It has 120B total and 12B active parameters.
What the release provides
The model supports image and video understanding with a context of up to one million tokens. Text, visual inputs, tool results and the response share that window. Multi-token prediction supports native speculative decoding.Nemotron 3.5 Super VL 120B-A12B official model card
How to use it well
Practical guidance. For a document assistant, keep the page image beside the extracted answer. Charts and screenshots need visual evidence that ordinary text extraction can omit. Reserve context for the response and tool results instead of filling the full window with input.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 |
| Release license | OpenMDW 1.1Nemotron 3.5 Super VL 120B-A12B official model card |