- MMLU-Pro 83.7
- AIME25 (no tools) 90.2
- HMMT Feb25 (no tools) 93.7
- GPQA (no tools) 79.2
- LiveCodeBench (v5 2024-07↔2024-12) 81.2
- SciCode (subtask) 42
- HLE (no tools) 18.3
- Terminal Bench (hard subset) 25.8
- Terminal Bench Core 2.0 31
- SWE-Bench (OpenHands) 60.5
- SWE-Bench Multilingual (OpenHands) 45.8
- TauBench V2 Average of Airline 56.25 / Retail 62.83 / Telecom 64.36 61.1
- BrowseComp with Search 31.3
- BIRD Bench 41.8
- IFBench (prompt) 72.6
- Scale AI Multi-Challenge 55.2
- Arena-Hard-V2 73.9
- AA-LCR 58.3
- RULER @ 1M 91.8
- MMLU-ProX (avg over langs) 79.4
- WMT24++ (en→xx) 86.7
Nemotron 3 Super
Nemotron 3 Super ranks first on LiveCodeBench v5.
42
Benchmarks with a score
1
Ranked first, out of five or more models
2
In the top three
NVIDIA Mar 2026 Nemotron 3 Super
Open weights
139.1
#184 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- IFBench 72.6% #32 /80
- SciCode 42% #25 /49
- AA-LCR (Artificial Analysis Long Context Reasoning) 58.3% #25 /38
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BullshitBench | 65.5% | #13 of 107 | Official leaderboard | |
| WMT24++ | 86.7% | #2 of 15 | Lab model card | |
| MMLU-Pro | 83.7% | #46 of 167 | Lab model card | |
| BullshitBench V2 | 54% | #45 of 129 | Official leaderboard | |
| RULER | 91.8% | #15 of 42 | Lab model card | |
| IFBench | 72.6% | #32 of 80 | Lab model card | |
| Arena-Hard v2.0 | 73.9% | #10 of 24 | Lab model card | |
| LMArena (Elo) | 1,360 | #169 of 335 | Official leaderboard | |
| MMLU-ProX | 79.4% | #12 of 23 | Lab model card | |
| MultiChallenge | 55.2% | #25 of 42 | Lab model card | |
| AA-LCR (Artificial Analysis Long Context Reasoning) | 58.3% | #25 of 38 | Lab model card | |
| Humanity's Last Exam | 18.3% | #111 of 154 | Lab model card | |
| SEA-HELM | 49.3 | #61 of 71 | Official leaderboard | |
| Extended NYT Connections | 15.4% | #92 of 103 | Official leaderboard | |
| StudyArena Text | 1,167 | #80 of 85 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HMMT 2025 | 93.7% | #17 of 54 | Lab model card | |
| AIME 2025 | 90.2% | #48 of 143 | Lab model card | |
| GPQA Diamond | 79.2% | #122 of 302 | Lab model card | |
| MathArena (Apex, IMO, USAMO) | 7.81% | #24 of 47 | Official leaderboard | |
| MathArena Apex Shortlist | 57.5% | #26 of 38 | Official leaderboard | |
| HMMT 2026 | 84.8% | #24 of 31 | Official leaderboard | |
| AIME 2026 | 91.7% | #27 of 31 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v5 | 81.2% | #1 of 47 | Lab model card | |
| LiveCodeBench v6 | 78.7% | #36 of 90 | Lab model card | |
| SciCode | 42% | #25 of 49 | Lab model card | |
| SWE-bench Verified | 60.5% | #109 of 155 | Lab model card | |
| ALE-Bench | 1,017 | #103 of 126 | Official leaderboard | |
| Bird-SQL | 41.8% | #11 of 13 | Lab model card | |
| OpenHands Index | 36.2% | #31 of 33 | Official leaderboard | |
| SWE-bench Multilingual | 45.8% | #69 of 73 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Terminal-Bench Hard | 25.8% | #6 of 13 | Lab model card | |
| tau2-bench | 61.1% | #65 of 79 | Lab model card | |
| PinchBench | 53.9% | #41 of 48 | Official leaderboard | |
| PinchBench 2.0 | 42.2% | #51 of 57 | Official leaderboard | |
| CivBench | 1,441 | #17 of 19 | Official leaderboard | |
| Terminal-Bench 2.0 | 31% | #59 of 63 | Lab model card | |
| BrowseComp | 31.3% | #76 of 81 | Lab model card | |
| Claw-Eval | 44.4% | #24 of 24 | Official leaderboard | |
| Claw-Eval Pass^3 | 5.50% | #24 of 24 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FoodTruck Bench | $962 | #48 of 59 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ABC Bench Liquid-Handling Robot | 47.7% | #29 of 34 | Official leaderboard | |
| ABC Bench Fragment Design | 74.4% | #24 of 27 | Official leaderboard |
Its lab's numbers put it at 139.1, number 184 of 404
The Self-Reported Index fits one score to the 22 capability results NVIDIA reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 137.0 to 141.2. How we compute it.
- HMMT 2025 93.7
- RULER 91.8
- AIME 2025 90.2
- WMT24++ 86.7
- MMLU-Pro 83.7
- LiveCodeBench v5 81.2
- MMLU-ProX 79.4
- GPQA Diamond 79.2
- LiveCodeBench v6 78.7
- Arena-Hard v2.0 73.9
- IFBench 72.6
- tau2-bench 61.1
- SWE-bench Verified 60.5
- AA-LCR (Artificial Analysis Long Context Reasoning) 58.3
- MultiChallenge 55.2
- SWE-bench Multilingual 45.8
- SciCode 42
- Bird-SQL 41.8
- BrowseComp 31.3
- Terminal-Bench 2.0 31
- Terminal-Bench Hard 25.8
- Humanity's Last Exam 18.3
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 1 of them isn't in the launch post.
Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning Mar 11, 2026
Capabilities
- MMLU-Pro 83.7
- AIME 2025 90.2
- HMMT 2025 93.7
- GPQA Diamond 79.2
- LiveCodeBench v5 81.2
- LiveCodeBench v6 78.7
- SciCode 42
- Humanity's Last Exam 18.3
- Terminal-Bench Hard 25.8
- Terminal-Bench 2.0 31
- SWE-bench Verified 60.5
- SWE-bench Multilingual 45.8
- tau2-bench 61.1
- BrowseComp 31.3
- Bird-SQL 41.8
- IFBench 72.6
- MultiChallenge 55.2
- Arena-Hard v2.0 73.9
- AA-LCR (Artificial Analysis Long Context Reasoning) 58.3
- RULER 91.6
- MMLU-ProX 79.4
- WMT24++ 86.7