NVIDIA-Nemotron-3-Super-120B-A12B-FP8: benchmark results

NVIDIA Nemotron 3 Super 120B A12B evaluated in FP8 quantization. Provider: NVIDIA. Released 2026-03-11. Access: Open.

Aggregate Rating 1689 ± 19, rank #223 of 1656 rated models, from 52 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Twinkle Eval - MMLU90.37Accuracy (%)100
Twinkle Eval - TMMLU+78.79Accuracy (%)100
RewardBench 2 Math90.57Score (%)98.4
MATH-MC Level 199.3Accuracy (%)97.8
RewardBench 2 Precise IF68.44Score (%)94.4
RewardBench 2 Factuality80Accuracy (%)91.5
MATH-MC Level 499.5Accuracy (%)88.2
NuclearQAv2 - Numeric74.11Accuracy (%) on 206 numeric calculation questions, correct w87.5
RewardBench 2 Focus91.11Score (%)87.1
Horangi 4 - Ko-AIME 202596.67Accuracy (%)87
MATH-MC Level 399.19Accuracy (%)86
Twinkle Eval - TW-Legal Benchmark v149.36Accuracy (%)85.7

Interactive version: theaggregate.ai/model?slug=nvidia-nemotron-3-super-120b-a12b-fp8 · How It Works · Data refreshed daily, snapshot 2026-10-20.