‹ PublicAI Index
The LLM benchmark aggregator.
NVIDIA Nemotron 3.5 Lightning 30B A3B
NVIDIA · 32B · A3B · open weights
Strongest in Cybersecurity (#45 of 47), weakest in Agents (#209 of 298). Above par in 2 of 17 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Fable 5.1.
§ 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 preference52−16.8#176/3541/1
Human preference52−16.8#176/3541/1
Reasoning38.7−26.7#177/1951/5
Mathematics32.9−30.9#151/1521/2
Coding34.9−31.2#179/1792/6
Agentic coding34.1−32.6#173/1732/5
Professional35.9−26.5#193/1941/2
Cybersecurity27−36.9#45/471/1
Medical32.9−33.6#143/1491/1
Finance37.6−24.8#146/1621/1
Legal35.5−30.4#165/1731/1
Core abilities32.8−32.8#200/2002/4
Instruction following26.8−47#60/621/1
Language26−40.6#62/621/1
General intelligence39.6−29.4#155/1801/3
Agents45.6−20.7#209/2982/6
Knowledge work43−27.7#135/2082/2
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for NVIDIA Nemotron 3.5 Lightning 30B A3B, left for the other.
§ 3 · Sources
Where the numbers come from
6 publications, 19 figures. Every one links to the page it was read from.
LMArena Text 1351
Artificial Analysis Intelligence Index 6
GDPval-AA 646
AA-Briefcase 507
LiveBench 1.0 44.5LiveBench 1.0 · Reasoning 42.9LiveBench 1.0 · Coding 45.8LiveBench 1.0 · Agentic Coding 19.9LiveBench 1.0 · Mathematics 66.5LiveBench 1.0 · Data Analysis 43.1LiveBench 1.0 · Language 38.3LiveBench 1.0 · Instruction Following 54.9
Vals · Legal Research Bench 2.88%Vals · Finance Agent 18.5%Vals · CorpFin 54.23%Vals · MedScribe 4.27%Vals · CyberBench 29.46%Vals · SWE-bench Verified 52.8% (Mini-SWE-agent)Vals · Code Migration 2.95%
Badge
[](https://publicai.io/model-index/m/nvidia-nemotron-3-5-lightning-30b-a3b)