NVIDIA AI & NIMopen-weight
Released: 2023-03-11

Nemotron 3 Super

NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer...

Intelligence Index
85.0/ 100
Calibrated multi-domain composite
Throughput Speed
85tok/sec
Output streaming throughput
Time To First Token
280ms
Initial chunk server latency
LiveBench Contam-Free
74.8%
September 2026 suite score

LiveBench Multi-Domain Evaluation Breakdown

Monthly refreshed contamination-free test sets
Logical Reasoning78.2%
Coding & Repo Repair76.5%
Mathematics (AIME)75.7%
Data Analysis & Tables72.3%
Language & Comprehension74.0%
Instruction Following74.8%

Sourced Benchmark Evaluations (61)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
LiveBench (Continuous)Active74.8%
94
2025-02-01independentSource ↗
ARC-AGI-3Active51.1%
93
2023-04-15independentSource ↗
ARC-AGI-2Nearing Saturation61.3%
75
2023-04-08independentSource ↗
FrontierMathActive21.1%
98
2023-04-05independentSource ↗
PutnamBenchActive39.2%
90
2023-04-04independentSource ↗
Omni-MATHDeprecated55.7%
10
2023-04-02independentSource ↗
SWE-LancerDeprecated52.4%
20
2023-04-02independentSource ↗
SoohakActive57.2%
98
2023-03-31independentSource ↗
FrontierCodeActive44.7%
98
2023-03-31independentSource ↗
Humanity's Last ExamActive34.6%
93
2023-03-29independentSource ↗
StructFlowBenchNearing Saturation60.6%
75
2023-03-29independentSource ↗
AIMENearing Saturation71.5%
75
2023-03-29independentSource ↗
OlympiadBenchActive61.7%
80
2023-03-28independentSource ↗
FireBenchActive59.9%
95
2023-03-27independentSource ↗
Terminal-BenchActive53.9%
80
2023-03-27independentSource ↗
MathArenaActive60.2%
99
2023-03-27independentSource ↗
OpenAI-MRCRActive68%
80
2023-03-26independentSource ↗
CodeContestsDeprecated50.1%
10
2023-03-26independentSource ↗
SimpleQAActive42.1%
90
2023-03-26independentSource ↗
ComplexBenchNearing Saturation58.5%
80
2023-03-25independentSource ↗
SWE-bench ProActive47.7%
90
2023-03-25independentSource ↗
LongBench-ProActive66.4%
80
2023-03-25independentSource ↗
MultiPL-EDeprecated70.8%
10
2023-03-24independentSource ↗
ZeroSCROLLSNearing Saturation64.8%
78
2023-03-24independentSource ↗
AgentIFNearing Saturation48.5%
80
2023-03-23independentSource ↗
MATHSaturated73.8%
25
2023-03-23independentSource ↗
LiveCodeBenchSaturated63.1%
39
2023-03-23independentSource ↗
InfiniteBenchNearing Saturation68.9%
65
2023-03-23independentSource ↗
MMLU-ProNearing Saturation74.5%
75
2023-03-22independentSource ↗
MercuryDeprecated67.8%
10
2023-03-22independentSource ↗
SCROLLSNearing Saturation64.8%
65
2023-03-22independentSource ↗
RULERNearing Saturation74.5%
84
2023-03-21independentSource ↗
TabMWPSaturated69.3%
25
2023-03-21independentSource ↗
CyberSecEvalDeprecated69.3%
10
2023-03-21independentSource ↗
MMLU-ReduxSaturated79.4%
25
2023-03-20independentSource ↗
QuALITYNearing Saturation71.3%
78
2023-03-20independentSource ↗
SWE-bench VerifiedSaturated60.1%
25
2023-03-20independentSource ↗
MGSMNearing Saturation71.5%
65
2023-03-19independentSource ↗
LongBenchNearing Saturation69.7%
65
2023-03-19independentSource ↗
SWE-benchDeprecated52.4%
10
2023-03-19independentSource ↗
AGIEvalSaturated67.6%
25
2023-03-18independentSource ↗
QASPERNearing Saturation67.2%
65
2023-03-18independentSource ↗
TruthfulQASaturated71.3%
25
2023-03-18independentSource ↗
SuperGLUESaturated80.2%
25
2023-03-17independentSource ↗
GSM-HardDeprecated72.3%
10
2023-03-17independentSource ↗
MT-BenchNearing Saturation7.3%
75
2023-03-16independentSource ↗
BIG-Bench HardSaturated77%
25
2023-03-16independentSource ↗
CodeXGLUEDeprecated72.4%
23
2023-03-16independentSource ↗
MBPP+Nearing Saturation78.5%
65
2023-03-15independentSource ↗
WinoGrandeSaturated81.7%
28
2023-03-15independentSource ↗
IFEvalDeprecated69.9%
25
2023-03-15independentSource ↗
GPQANearing Saturation71.3%
78
2023-03-15independentSource ↗
HumanEval+Saturated83.2%
25
2023-03-14independentSource ↗
Needle-in-a-HaystackSaturated96.2%
44
2023-03-14independentSource ↗
CommonsenseQASaturated82.5%
28
2023-03-14independentSource ↗
GSM8KSaturated81.3%
25
2023-03-13independentSource ↗
GLUESaturated84.9%
28
2023-03-13independentSource ↗
MBPPSaturated81.6%
25
2023-03-13independentSource ↗
HumanEvalSaturated87.8%
25
2023-03-12independentSource ↗
MMLUSaturated82.6%
25
2023-03-12independentSource ↗
ARC-AGISaturated73.9%
25
2023-03-12independentSource ↗
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run nvidia-nemotron-3-super

# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nvidia-nemotron-3-super",
    "messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
  }'

Evaluation & Sourcing Notes

Scores listed for Nemotron 3 Super represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.

All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.