Inference
EN

Nemotron 3.5 Lightning 30B A3B NVFP4 vs Nemotron 3.5 Super VL 120B A12B

Change models
Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA
Nemotron 3.5 Super VL 120B A12BNVIDIA
IndexAll categories, out of 100.
37.0 ±6.8
51.0 ±5.1 best
GapIndex points below the top.
−14.0
Top
CoverageIndex weight that has results.
75%
85% best
SpeedOutput tokens per second.
—
—
EloLMArena rating and rank.
—
—
ReleasedMonth of release.
2026-08
2026-10
Input / 1MUS dollars per 1M input tokens.
—
—
Output / 1MUS dollars per 1M output tokens.
—
—
1K chats2K tokens in, 500 out, each.
—
—
1K documents20K tokens in, 1K out, each.
—
—
1K agent tasks200K tokens in, 20K out, each.
—
—
ContextMaximum tokens in one request.
1M
1M
AgenticMulti-step tasks with tools.
30.2
47.8 best
CodingCode writing and repair.
38.7
53.0 best
ReasoningLogic problems and puzzles.
38.9
56.1 best
MultimodalTasks with images and text.
—
47.4
KnowledgeFacts and expert knowledge.
39.5
52.7 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
57.1 best
55.2
MathMath problems.
—
—

Benchmarks

13 shared
Only Nemotron 3.5 Lightning 30B A3B NVFP4: 6Only Nemotron 3.5 Super VL 120B A12B: 15
Nemotron 3.5 Lightning 30B A3B NVFP4
Nemotron 3.5 Super VL 120B A12B
Artificial Analysis Humanity's Last ExamKnowledge
10.6%143 / 222
30.7% best74 / 222
Artificial Analysis Long Context ReasoningReasoning
49.2%159 / 217
68.6% best121 / 217
Critical Physics TasksReasoning
0.0%150 / 214
10.9% best62 / 214
Artificial Analysis Omniscience AccuracyKnowledge
14.7%153 / 164
26.2% best97 / 164
GDPval-AA normalizedAgentic
6.9%93 / 133
33.1% best55 / 133
Artificial Analysis SciCodeCoding
32.1%102 / 114
47.8% best61 / 114
Software Engineering Benchmark VerifiedCoding
52.8%94 / 114
71.1% best64 / 114
Terminal-Bench 2.1 · BenchLMAgentic
23.5%82 / 86
66.7% best56 / 86
GPQA Diamond · BenchLMKnowledge
75.6%68 / 84
89.5% best33 / 84
BrowseCompAgentic
36.8%62 / 63
78.0% best31 / 63
Instruction Following BenchmarkInstruction
72.9% best37 / 57
70.9%40 / 57
Humanity's Last Exam without toolsKnowledge
10.5%51 / 53
31.8% best35 / 53
τ³-Bench Tool-Agent-User EvaluationAgentic
9.5%30 / 32
18.9% best26 / 32