BlogLLM Benchmark Scores as Published, Side by Side
Nemotron 3 Ultra: Benchmark Scores as Published
Every result the catalog holds for Nemotron 3 Ultra: 64 tests in 14 areas, from 5 pages. 46 printed by NVIDIA itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer NVIDIAReleased 2026-06-04Context window 1MInput TextOutput TextWeights Open (OpenMDW License Agreement, version 1.1)Data version 2026103001, as of 2026-10-30
Knowledge Work
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AutomationBench | 23.28% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 1 more page | - |
| GDPval-AA v2 | 1148 | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
| APEX-Agents | 9% | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| JobBench | 27.2 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | - |
| GDPval-AA (Elo) | 1162 | IFM, IFM K2-Horizon-375B-A23B model cardchecked 2026-10-15 | - |
| Vals Finance Agent v1.1 | 60.1% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Tools: none |
| 53.7% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Tools: web search | |
| GDPval (metric not stated) | 46.7% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | - |
| ProfBench (Search) | 56% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Tools: web search |
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 73.9% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
| SWE-Bench Pro | 49.66% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
| Terminal-Bench 2.1 | 54.8% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 4 more pages | - |
| DeepSWE v1.1 | 3% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19 | - |
| SWE-bench Multilingual | 68.8% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 1 more page | - |
| Terminal-Bench 4.0 | 0% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19 | - |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 87% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 3 more pages | - |
| SciCode | 42% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 3 more pages | - |
| IMO-AnswerBench | 92.3% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Tools: with tools |
| 88.6% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Tools: none | |
| AIME 2026 | 94.2% | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | - |
| CritPt | 3.1% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
| MathArena Apex Shortlist | 75% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 1 more page | - |
| GPQA | 87% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Tools: none |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AA-LCR | 75.2% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
| LongBench v2 | 61.9% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | - |
| GraphWalks (variant not stated) | 63.3% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19 | - |
| RULER at 1M | 94.7% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | Context: 1048576 |
Multimodal Understanding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GDP.pdf | 8.4% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19 | - |
General Knowledge and Reasoning
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 89.2% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 1 more page | - |
| IOI 2025 (total points) | 570 | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | - |
Tool Use and Workflows
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| BrowseComp | 67.85% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| IFBench | 82.9 | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 2 more pages | - |
| MultiChallenge | 63.8% | NVIDIA, NVIDIA Nemotron 3 Ultra 550B-A55B model cardchecked 2026-10-15 | - |
Multilingual
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMLU-ProX | 83.1% | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 1 more page | - |
| WMT24++ (XCOMET-XXL) | 82.7 | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19; printed on 1 more page | - |
| Global-MMLU-Lite | 85.6% | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | - |
Writing and Conversation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LMArena v3 proxy | 34.3 | NVIDIA, NVIDIA Nemotron 3.5 Super VL 120B-A12B model cardchecked 2026-10-19 | - |
Safety and Alignment
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| FORTRESS (Adversarial) | 77.6% | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | - |
| FORTRESS (Benign) | 90.5% | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | - |
| StrongREJECT | 98.7% | Thinking Machines, Thinking Machines Inkling model cardchecked 2026-10-15 | - |
| FORTRESS adversarial (static), ARSlower is better | 26.4 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | - |
| FORTRESS adversarial (static), ORSlower is better | 13.3 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | - |
| StrongREJECT v2 (adaptive)lower is better | 80.7 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Artificial Analysis Intelligence Index | 23 | Meta, Meta Muse Spark Open 816B-A42B model cardchecked 2026-10-19 | - |
More from NVIDIA
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.