Nemotron 3 Ultra: benchmark results
NVIDIA's flagship open Nemotron 3 Ultra MoE model (550B total, 55B active). Provider: NVIDIA. Released 2026-06-04. Access: Open.
Aggregate Rating 1681 ± 7, rank #240 of 1635 rated models, from 346 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats (RULER) | 94.7 | Score (%) | 100 |
| PinchBench - 147 Tasks - OpenClaw Version Unspecified | 90.58 | Best Official Run Score (%) | 100 |
| AA IFBench | 81.36 | Accuracy (%) | 99.3 |
| PACT - Rule-Scope Discernment | 0.89 | Score (0-1) | 95.2 |
| FinRiskAtlas - Evidence-Grounded Processing - Quantitative Reasoning | 81.9 | Numerical Accuracy (%) | 93.8 |
| AA-Omniscience Hallucination Rate | 29.69 | Hallucination Rate (%) | 93 |
| IKP - T2 | 100 | Accuracy (%) | 92.6 |
| Dnotitia Korean LLM Leaderboard | 97.18 | Mean Item Score (%) | 89.8 |
| FinRiskAtlas - Applied Review - Risk Classification | 75.41 | Accuracy (%) | 89.1 |
| Enkrypt AI - Insecure Code Risk | 6.22 | Risk Score | 88.7 |
| FinRiskAtlas - Domain Knowledge - Finance & Law | 81.85 | Macro-Averaged Accuracy (%) | 87.5 |
| Arcee MMLU-Pro (BenchLM) | 86.8 | Accuracy (%) | 86.9 |
Interactive version: theaggregate.ai/model?slug=nemotron-3-ultra · How It Works · Data refreshed daily, snapshot 2026-10-15.