NVIDIA · model
Nemotron 3 Ultra benchmark scores
As of 2026-11-21, Nemotron 3 Ultra (NVIDIA) has measured scores on 40 benchmarks and estimated scores on 121 more.
| Benchmark | Score | Source |
|---|---|---|
| AA-GPQA Diamond | 86.7% | measured |
| AA-HLE | 28.4% | measured |
| AA-SciCode | 40.3% | measured |
| Artificial Analysis Intelligence Index | 11.8% | measured |
| CritPt | 3.1% | measured |
| GDPval-AA | 33.1% | measured |
| GPQA Diamond | 87.0% | measured |
| HLE | 26.7% | measured |
| IFBench | 81.7% | measured |
| AA-LCR | 67.0% | measured |
| LiveCodeBench v6 | 89.0% | measured |
| AA-Omniscience Accuracy | 21.6% | measured |
| SciCode | 44.6% | measured |
| Terminal-Bench 2.1 | 56.4% | measured |
| BrowseComp | 44.4% | measured |
| HLE w/ tools | 37.4% | measured |
| IFEval | 93.2% | estimated ± 2.8 pp, medium confidence |
| LongBench v2 | 61.9% | measured |
| VITA-Bench | 12.7% | estimated ± 6.7 pp, low confidence |
| τ²-bench results | 83.3% | measured |
| AA Coding Index | 49.3% | measured |
| APEX-Agents | 25.2% | estimated ± 9.4 pp, low confidence |
| APEX-Agents-AA | 7.4% | estimated ± 4.4 pp, medium confidence |
| BioMysteryBench (human-difficult) | 31.4% | estimated ± 6.8 pp, low confidence |
| DeepSearchQA | 85.4% | estimated ± 4.5 pp, high confidence |
| SWE-bench Verified | 71.9% | measured |
| AutomationBench | 48.6% | estimated ± 9.1 pp, low confidence |
| BFCL v4 | 74.9% | estimated ± 2.7 pp, medium confidence |
| CyberGym | 62.2% | estimated ± 5.3 pp, low confidence |
| JobBench | 32.3% | estimated ± 8.7 pp, medium confidence |
| SWE-bench Pro | 49.5% | estimated ± 4.5 pp, high confidence |
| τ³-bench results | 70.9% | measured |
| WideResearch | 55.9% | estimated ± 3.5 pp, medium confidence |
| AA Agentic Index | 21.7% | measured |
| MMLU-Pro | 86.8% | measured |
| AA-IFBench | 81.4% | measured |
| BBH | 99.8% | estimated ± 2.8 pp, medium confidence |
| FailureSensorIQ | 66.5% | estimated ± 5.0 pp, medium confidence |
| GPQA | 87.0% | measured |
| MMLU | 89.9% | estimated ± 0.9 pp, high confidence |
| Vals GPQA Diamond | 86.1% | measured |
| Vals LiveCodeBench | 86.0% | measured |
| Vals MMLU-Pro | 85.8% | measured |
| MMMLU | 87.7% | estimated ± 0.7 pp, medium confidence |
| Chess Puzzles | 36.1% | estimated ± 4.7 pp, medium confidence |
| AA MMLU-Pro | 84.6% | estimated ± 0.7 pp, medium confidence |
| Gert Labs | 26.8% | estimated ± 7.1 pp, low confidence |
| AA-AnalystAgent | 44.8% | estimated ± 5.9 pp, low confidence |
| AA EnterpriseOps-Gym | 35.8% | estimated ± 7.2 pp, low confidence |
| AA Harvey LAB v1.0 | 81.7% | measured |
| AA Tau3 Banking | 47.1% | estimated ± 2.4 pp, low confidence |
| AA Terminal-Bench 2.1 | 79.8% | estimated ± 2.5 pp, medium confidence |
| ApprenticeBench | 13.1% | estimated ± 12.5 pp, low confidence |
| ARC-AGI-1 | 84.4% | estimated ± 2.5 pp, medium confidence |
| ARC-AGI-2 | 40.6% | estimated ± 9.7 pp, low confidence |
| CursorBench 3.1 | 47.9% | estimated ± 4.8 pp, low confidence |
| CursorBench 3.2 | 34.2% | estimated ± 4.4 pp, medium confidence |
| EBR-bench | 8.9% | estimated ± 10.4 pp, low confidence |
| Terminal-Bench 3.0 | 2.2% | estimated ± 6.9 pp, low confidence |
| FrontierCode 1.1 Main | 13.2% | estimated ± 4.2 pp, low confidence |
| FrontierSWE v2 | 6.1% | estimated ± 9.5 pp, low confidence |
| Furniture Assembly | 33.0% | estimated ± 9.3 pp, medium confidence |
| MaxxBench | 34.6% | estimated ± 5.7 pp, low confidence |
| MazeBench | 0.0% | measured |
| Mystery Game Puzzles | 20.0% | measured |
| NYT Connections | 62.6% | estimated ± 13.8 pp, low confidence |
| OSWorld-Verified | 50.3% | estimated ± 5.7 pp, low confidence |
| Terminal-Bench 4.1 | 0.0% | estimated ± 10.9 pp, low confidence |
| Terminal-Bench Hard | 45.1% | estimated ± 3.9 pp, medium confidence |
| Vals SWE-bench | 69.0% | measured |
| Terminal-Bench 2.1 (Vals AI) | 50.9% | measured |
| VulcanBench v3 | 84.3% | estimated ± 5.6 pp, low confidence |
| CWE-Bench-AA | 27.5% | measured |
| AA Cyber Index | 13.1% | measured |
| DeepsecBench-AA | 5.8% | estimated ± 8.9 pp, low confidence |
| HealthBench Professional (length-adjusted, Artificial Analysis) | 49.4% | estimated ± 4.3 pp, low confidence |
| HealthBench Professional (raw, Artificial Analysis) | 32.3% | estimated ± 0.8 pp, low confidence |
| AA ITBench | 39.8% | estimated ± 4.0 pp, high confidence |
| Bug Hunt Bench | 14.8% | estimated ± 10.5 pp, low confidence |
| CursorBench 4.0 | 30.8% | estimated ± 3.4 pp, medium confidence |
| CWE-bench v1 | 57.4% | estimated ± 5.1 pp, medium confidence |
| DeepSWE | 56.3% | estimated ± 7.2 pp, low confidence |
| DRACO | 86.9% | estimated ± 0.7 pp, low confidence |
| FrontierCode 2.0 | 26.1% | estimated ± 9.3 pp, low confidence |
| FrontierPhysics | 1.0% | estimated ± 4.7 pp, medium confidence |
| HLE w/o tools | 26.7% | measured |
| OSWorld 2.0 | 1.8% | estimated ± 14.7 pp, low confidence |
| OSWorld 2.1 | 57.3% | estimated ± 5.3 pp, low confidence |
| OSWorld 2.1 Offline | 11.7% | estimated ± 1.9 pp, low confidence |
| PostTrainBench v1.1 | 14.5% | estimated ± 3.7 pp, medium confidence |
| SpireClimb | 0.0% | estimated ± 9.3 pp, medium confidence |
| StealthMark | 16.7% | measured |
| SWE Multilingual | 67.7% | measured |
| Toolathlon-Verified | 50.4% | estimated ± 2.3 pp, low confidence |
| VulcanBench Frontier v4 | 70.1% | estimated ± 6.8 pp, low confidence |
| AA AutomationBench | 41.9% | estimated ± 7.9 pp, low confidence |
| GDP.pdf | 17.0% | estimated ± 4.6 pp, medium confidence |
| AA Terminal-Bench 4.0 | 4.3% | estimated ± 11.0 pp, low confidence |
| ARC-AGI-3 | 0.3% | estimated ± 7.2 pp, low confidence |
| ARC-AGI-3 Provider Adapter | 2.7% | estimated ± 7.3 pp, low confidence |
| CWE-bench | 53.9% | estimated ± 7.9 pp, low confidence |
| MKA | 41.9% | estimated ± 5.7 pp, low confidence |
| SkillsBench official leaderboard (BenchFlow) | 50.1% | estimated ± 6.3 pp, low confidence |
| CompileBench | 84.4% | estimated ± 10.0 pp, low confidence |
| OTelBench | 22.6% | estimated ± 4.1 pp, medium confidence |
| Vibe Code Bench | 0.2% | estimated ± 8.4 pp, medium confidence |
| FACTS Multimodal | 39.3% | estimated ± 4.7 pp, medium confidence |
| Hermes Bench | 62.0% | estimated ± 2.9 pp, high confidence |
| Hermes Index | 37.2% | estimated ± 3.1 pp, high confidence |
| ExploitBench | 36.7% | estimated ± 7.2 pp, low confidence |
| ExploitGym | 8.8% | estimated ± 2.7 pp, medium confidence |
| C-Eval | 92.3% | estimated ± 0.9 pp, medium confidence |
| Claw-Eval | 62.7% | estimated ± 6.4 pp, low confidence |
| DeepPlanning | 35.4% | estimated ± 8.3 pp, low confidence |
| INCLUDE | 83.7% | estimated ± 1.4 pp, low confidence |
| MAXIFE | 77.3% | estimated ± 6.5 pp, low confidence |
| MCP Atlas | 65.7% | estimated ± 7.8 pp, low confidence |
| MCP-Tasks | 75.1% | estimated ± 3.6 pp, low confidence |
| MMLU-ProX | 83.0% | measured |
| MMLU-Redux | 94.1% | estimated ± 0.6 pp, high confidence |
| NL2Repo | 42.6% | estimated ± 3.4 pp, high confidence |
| NOVA-63 | 56.7% | estimated ± 1.3 pp, low confidence |
| PolyMath | 49.0% | estimated ± 11.4 pp, low confidence |
| PostTrainBench v1 | 10.7% | estimated ± 3.6 pp, medium confidence |
| QwenClawBench | 86.4% | estimated ± 6.0 pp, low confidence |
| SuperGPQA | 67.4% | estimated ± 7.2 pp, medium confidence |
| Terminal-Bench 2.0 | 59.4% | estimated ± 5.8 pp, medium confidence |
| Toolathlon | 21.2% | estimated ± 4.2 pp, medium confidence |
| AA LiveCodeBench | 86.3% | estimated ± 2.6 pp, high confidence |
| HealthBench Hard | 0.0% | estimated ± 5.1 pp, low confidence |
| MedXpertQA (Text) | 45.9% | estimated ± 4.5 pp, low confidence |
| MMLU-Pro (Arcee) | 85.8% | estimated ± 3.4 pp, medium confidence |
| React Native Evals | 76.0% | estimated ± 2.7 pp, medium confidence |
| ResearchClawBench | 16.9% | estimated ± 1.9 pp, high confidence |
| SWE-Rebench | 57.0% | estimated ± 3.6 pp, medium confidence |
| PostTrainBench | 12.1% | estimated ± 3.7 pp, medium confidence |
| DAYJOB | 0.0% | estimated ± 3.2 pp, medium confidence |
| DAYJOB Healthcare | 0.7% | estimated ± 4.2 pp, high confidence |
| Finance Agent v2 | 56.1% | estimated ± 3.8 pp, medium confidence |
| BioMysteryBench (human-solvable) | 81.7% | estimated ± 0.6 pp, low confidence |
| HealthBench (raw) | 37.2% | estimated ± 6.0 pp, low confidence |
| HealthBench (length-adjusted) | 49.2% | estimated ± 3.4 pp, low confidence |
| HealthBench Professional | 36.6% | estimated ± 4.2 pp, low confidence |
| HealthBench Professional (raw) | 48.5% | estimated ± 4.5 pp, medium confidence |
| HLE-Verified | 2.3% | estimated ± 4.6 pp, low confidence |
| LABBench2 | 80.8% | estimated ± 1.8 pp, low confidence |
| ScarfBench | 24.6% | estimated ± 9.1 pp, low confidence |
| BlenderWorld (text-only) | 70.1% | estimated ± 2.8 pp, medium confidence |
| Cookie-Bench | 77.1% | estimated ± 1.3 pp, medium confidence |
| STEMBench | 2.3% | estimated ± 0.2 pp, low confidence |
| SimpleQA | 20.3% | estimated ± 4.3 pp, medium confidence |
| LiveCodeBench | 74.2% | estimated ± 14.6 pp, low confidence |
| Agents' Last Exam | 35.4% | estimated ± 11.0 pp, low confidence |
| OpenHarmony Bench v1.0 | 51.0% | estimated ± 2.4 pp, medium confidence |
| SEC-Bench Pro | 17.2% | estimated ± 9.3 pp, low confidence |
| MRCRv2 | 58.0% | estimated ± 11.3 pp, low confidence |
| MMLongBench-Doc | 60.8% | estimated ± 2.6 pp, low confidence |
| BioBasicsBench | 33.9% | estimated ± 2.8 pp, low confidence |
| LifeVal | 9.3% | estimated ± 1.6 pp, low confidence |
| ExploitBench Internal Port | 0.1% | estimated ± 5.2 pp, low confidence |
| PinchBench | 90.0% | measured |