The LLM benchmark aggregator.

NVIDIA Nemotron 3 Super 120B A12B

NVIDIA · 124B · A12B · open weights

Strongest in Secure code (#40 of 268), weakest in Jailbreak resistance (#245 of 266). Above par in 4 of 21 scopes.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Human preference52.8−16#168/3541/1

Agents47.2−19.1#173/2981/6

Knowledge work44.5−26.2#126/2081/2

Tool use42.1−31.9—/81✱0/1

Safety47.5−13.9#236/3271/3

Secure code62.3−3.9#40/2681/1

Toxicity avoidance56.3−2.7#59/2661/1

Fairness45.2−24.8#186/2901/2

Harm refusal45.1−16.4#237/2901/2

Jailbreak resistance31.9−33.1#245/2661/1

Core abilities47.3−18.3—/200✱0/4

Long context40−24.6—/0✱✱

Knowledge48.2−14.3—/197✱0/3

Factuality41.4−19.4—/0✱✱

Reasoning49−16.4—/195✱0/5

Science49.6−13.9—/122✱0/1

Expert reasoning40.9−25.5—/0✱✱

Coding49.7−16.4—/179✱0/6

Agentic coding49.5−17.2—/173✱0/5

Code generation49.4−15.9—/20✱0/1

§ 3 · Sources

Where the numbers come from

4 publications, 17 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-11

LMArena Text 1360

GDPval-AA ↗1 measureread 2026-11-11

GDPval-AA 497

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-11-11

Enkrypt · Jailbreak risk 28.1%Enkrypt · Harmful content risk 7.8%Enkrypt · CBRN risk 49.2%Enkrypt · Toxicity risk 1.8%Enkrypt · Bias risk 85.3%Enkrypt · Insecure code risk 8%

✱✱ IFM — K2 Horizon launch ↗the model’s own publisher9 measuresread 2026-11-03

tau3-Banking 10.3%Terminal-Bench 2.1 38.6%SciCode 36%Humanity's Last Exam (without tools) 20.8%GPQA Diamond 80%CritPt 3.1%AA-LCR 60.3%AA-Omniscience Accuracy 24.3%AA-Omniscience Non-Hallucination 13%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=nvidia-nemotron-3-super-120b-a12b)](https://publicai.io/model-index/m/nvidia-nemotron-3-super-120b-a12b)