Nemotron 3.5 Super VL (120B-A12B, NVFP4)
NVIDIA's flagship Nemotron 3.5 Super VL at 120B parameters with 12B active, in NVIDIA's quantization-aware NVFP4. Native reasoning, tool calling, image understanding, and a 256K context window, with its own multi-token-prediction head for faster replies. Workstation-tier: needs about 90GB of usable memory, and runs best on NVIDIA Blackwell GPUs such as DGX Spark; not offered on Macs.
| At a glance | |
|---|---|
| Made by | NVIDIA |
| Download size | 76.9 GB |
| Parameters | 120B |
| Quantization | NVFP4 |
| Context window | 262K tokens |
| Uses tools | Yes |
| Best for | Reasoning |
| Runs on | llama.cpp |
| License | OpenMDW 1.1 |
| Added to the catalog | 2026-10-17 |
Gezel downloads this model the first time a gezel needs it, checks the file against its published hash, and applies tuned settings for it.