Models

NVIDIA Nemotron 3 Super 120B A12B BF16

NVIDIA · March 2026

Get

Hugging Face

A 121B, 2.4B active chat model from NVIDIA.

Rating: 3.2
Downloads: 2.3K
Parameters: 121B, 2.4B active
Context: 1M
Licence: other

Too big for this Mac

Its smallest size needs 59 GB of memory, more than M4 · 16 GB can give it.

Ratings

3.2 out of 5, across the tests below.

KnowledgeMMLU-Pro: 83.7%
ScienceGPQA Diamond: 82.7%
Expert questionsHumanity's Last Exam: 22.8%
Software fixesSWE-bench Verified: 60.5%
Terminal tasksTerminal-Bench 2.0: 31%
MathsAIME 2026: 90%

MMLU-Pro: graduate-level questions in 14 subjects. GPQA Diamond: questions PhD scientists write for each other. Humanity's Last Exam: questions written to be beyond today's models. SWE-bench Verified: real GitHub issues fixed so the project's tests pass. Terminal-Bench 2.0: jobs finished alone in a terminal. AIME 2026: this year's AIME competition problems. Scores are the ones filed on the model's Hugging Face page or the board's own table, read 2026-11-05; some come from the maker.

NVIDIA Nemotron 3 Super 120B A12B BF16 runs well from a Mac with an M2 Max and 96 GB of memory. 15 of 59 Mac configurations run it well.

Compare with

The nearest models in size.

Sizes

From mradermacher/NVIDIA-Nemotron-3-Super-120B-A12B-BF16-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.

Size	Download	Memory at 8K	Get
Q2_K	54 GB	54 GBestimated	Open
Q3_K_S	60 GB	60 GBestimated	Open
Q3_K_M	67 GB	67 GBestimated	Open
IQ4_XS	68 GB	68 GBestimated	Open
Q3_K_L	71 GB	71 GBestimated	Open
Q4_K_S	76 GB	76 GBestimated	Open
Q4_K_M	86 GB	86 GBestimated	Open
Q5_K_S	87 GB	87 GBestimated	Open
Q5_K_M	96 GB	96 GBestimated	Open
Q6_K	113 GB	113 GBestimated	Open
Q8_0	128 GB	129 GBestimated	Open

Run It Anywhere

NVIDIA Nemotron 3 Super 120B A12B BF16 is an open model, so any app that runs GGUF files can run it. These take the Q2_K size from mradermacher straight from Hugging Face.

LM StudioMac, Windows, LinuxOpen
JanMac, Windows, LinuxOpen
MstyMac, Windows, LinuxOpen
OllamaMac, Windows, Linuxollama run hf.co/mradermacher/NVIDIA-Nemotron-3-Super-120B-A12B-BF16-GGUF:Q2_K
llama.cppMac, Windows, Linuxllama-server -hf mradermacher/NVIDIA-Nemotron-3-Super-120B-A12B-BF16-GGUF:Q2_K
Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/mradermacher/NVIDIA-Nemotron-3-Super-120B-A12B-BF16-GGUF:Q2_K

On a PC Graphics Card

Video memory	Largest size that fits	Memory at 8K
6 GB	None worth running	needs 60 GB
8 GB	None worth running	needs 60 GB
10 GB	None worth running	needs 60 GB
11 GB	None worth running	needs 60 GB
12 GB	None worth running	needs 60 GB
16 GB	None worth running	needs 60 GB
20 GB	None worth running	needs 60 GB
24 GB	None worth running	needs 60 GB
32 GB	None worth running	needs 60 GB
48 GB	None worth running	needs 60 GB

The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.

On every Mac

Mac	Size	Fit	Memory	Writes
M6 · 16 GB	Q2_K	Too big	59 GB
M6 · 24 GB	Q2_K	Too big	59 GB
M6 · 32 GB	Q2_K	Too big	59 GB
M5 · 16 GB	Q2_K	Too big	59 GB
M5 · 24 GB	Q2_K	Too big	59 GB
M5 · 32 GB	Q2_K	Too big	59 GB
M5 Pro · 24 GB	Q2_K	Too big	59 GB
M5 Pro · 48 GB	Q2_K	Too big	59 GB
M5 Pro · 64 GB	Q2_K	Too big	59 GB
M5 Max · 36 GB	Q2_K	Too big	59 GB
M5 Max · 48 GB	Q2_K	Too big	59 GB
M5 Max · 64 GB	Q2_K	Too big	59 GB
M5 Max · 128 GB	IQ4_XS	Runs great	68 GB	71
M5 Ultra · 96 GB	IQ4_XS	Runs well	68 GB	92
M5 Ultra · 256 GB	Q8_0	Runs great	134 GB	87
M5 Ultra · 512 GB	Q8_0	Runs great	134 GB	87
M4 · 16 GB	Q2_K	Too big	59 GB
M4 · 24 GB	Q2_K	Too big	59 GB
M4 · 32 GB	Q2_K	Too big	59 GB
M4 Pro · 24 GB	Q2_K	Too big	59 GB
M4 Pro · 48 GB	Q2_K	Too big	59 GB
M4 Pro · 64 GB	Q2_K	Too big	59 GB
M4 Max · 36 GB	Q2_K	Too big	59 GB
M4 Max · 48 GB	Q2_K	Too big	59 GB
M4 Max · 64 GB	Q2_K	Too big	59 GB
M4 Max · 96 GB	IQ4_XS	Runs well	68 GB	25
M4 Max · 128 GB	IQ4_XS	Runs great	68 GB	27
M3 · 8 GB	Q2_K	Too big	59 GB
M3 · 16 GB	Q2_K	Too big	59 GB
M3 · 24 GB	Q2_K	Too big	59 GB
M3 Pro · 18 GB	Q2_K	Too big	59 GB
M3 Pro · 36 GB	Q2_K	Too big	59 GB
M3 Max · 36 GB	Q2_K	Too big	59 GB
M3 Max · 48 GB	Q2_K	Too big	59 GB
M3 Max · 64 GB	Q2_K	Too big	59 GB
M3 Max · 96 GB	IQ4_XS	Runs well	68 GB	26
M3 Max · 128 GB	IQ4_XS	Runs great	68 GB	29
M3 Ultra · 96 GB	IQ4_XS	Runs well	68 GB	33
M3 Ultra · 256 GB	Q8_0	Runs great	134 GB	34
M3 Ultra · 512 GB	Q8_0	Runs great	134 GB	34
M2 · 8 GB	Q2_K	Too big	59 GB
M2 · 16 GB	Q2_K	Too big	59 GB
M2 · 24 GB	Q2_K	Too big	59 GB
M2 Pro · 16 GB	Q2_K	Too big	59 GB
M2 Pro · 32 GB	Q2_K	Too big	59 GB
M2 Max · 32 GB	Q2_K	Too big	59 GB
M2 Max · 64 GB	Q2_K	Too big	59 GB
M2 Max · 96 GB	IQ4_XS	Runs well	68 GB	28
M2 Ultra · 64 GB	Q2_K	Too big	59 GB
M2 Ultra · 128 GB	IQ4_XS	Runs great	68 GB	33
M2 Ultra · 192 GB	Q8_0	Runs great	131 GB	33
M1 · 8 GB	Q2_K	Too big	59 GB
M1 · 16 GB	Q2_K	Too big	59 GB
M1 Pro · 16 GB	Q2_K	Too big	59 GB
M1 Pro · 32 GB	Q2_K	Too big	59 GB
M1 Max · 32 GB	Q2_K	Too big	59 GB
M1 Max · 64 GB	Q2_K	Too big	59 GB
M1 Ultra · 64 GB	Q2_K	Too big	59 GB
M1 Ultra · 128 GB	IQ4_XS	Runs great	68 GB	21

Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.

Information

Maker: NVIDIA
Released: March 2026
Parameters: 121B, 2.4B active
Longest chat: 1M tokens
Reads pictures: No
Licence: other
Downloads this month: 2,292
Sizes: 11, 54 to 128 GB
Kind: Maker's release
Source: nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16

Badge for the model card

Shows on NVIDIA Nemotron 3 Super 120B A12B BF16's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Files, downloads and likes read from Hugging Face on November 5, 2026 and refreshed every morning. Memory and speed are estimates until someone running Outmake measures this model on that Mac.