---

base_model:

- nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-LiLiCorr

license: other

license_name: openmdw-1.1

license_link: https://openmdw.ai/license/1-1/

library_name: llama.cpp

tags:

- gguf

- nvfp4

- lilicorr

- speculative-decoding

- magnitude

inference: false

---

NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 LiLiCorr GGUF

Native NVFP4 GGUF conversion of

nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-LiLiCorr

for use with Magnitude.

Converted with

llama.cpp

at revision 620d2225cc436ab0b001e3200e590e3b286ef9ab.

Artifact

• File: NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-LiLiCorr.gguf
• Quantization: NVFP4
• SHA-256: 7123a836be50b56fa512c9da5abe560b4b9090df1353a089a8935356a7392401
• Source revision: 7c7221aeb4cb3139307aa457f7dce11a8fb6ceb2
• Target tokenizer revision: a1ccf94d69377bf82c4db39798aac28447d9c16f

Draft metadata

• dflash.attention.causal = false: the draft block uses bidirectional attention.
• dflash.sample_from_anchor = false: proposals come from mask rows, starting after the anchor.
• dflash.attention.sliding_window = 8192 on every layer.
• Derived from the NVIDIA checkpoint configuration.