- MMLU-Pro 78.3
- AIME25 (no tools) 89.1
- GPQA (no tools) 73
- LiveCodeBench (v6 2025-08–2025-05) 68.3
- SciCode (subtask) 33.3
- HLE (no tools) 10.6
- MiniF2F pass@1 50
- Terminal Bench (hard subset) 8.5
- SWE-Bench (OpenHands) 38.8
- TauBench V2 (Average) of Airline 48.0 / Retail 56.9 / Telecom 42.2 49
- BFCL v4 53.8
- IFBench (prompt) 71.5
- Scale AI Multi Challenge 38.5
- Arena-Hard-V2 (Average); Hard Prompt 72.1, Creative Writing 63.2 67.7
- AA-LCR 35.9
- RULER-100@1M 86.3
- MMLU-ProX (avg over langs) 59.5
- WMT24++ (en->xx) 86.2
Nemotron 3 Nano 30B-A3B
Nemotron 3 Nano 30B-A3B places in the top three on WMT24++.
27
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
NVIDIA Dec 2025 Nemotron 3 Nano 30B-A3B
Open weights Small
No Epoch Capabilities Index score yet
- IFBench 71.5% #37 /70
- SciCode 33.3% #40 /47
- AA-LCR (Artificial Analysis Long Context Reasoning) 35.9% #30 /31
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 | |
|---|---|---|---|---|
| WMT24++ | 86.2% | #3 of 12 | Model card | |
| BullshitBench | 27.3% | #52 of 107 | Official leaderboard | |
| IFBench | 71.5% | #37 of 70 | Model card | |
| Vectara Hallucination Leaderboard (2025 dataset) | 9.60% | #56 of 104 | Official leaderboard | |
| MMLU-Pro | 78.3% | #85 of 151 | Model card | |
| Arena-Hard v2.0 | 67.7% | #14 of 24 | Model card | |
| BullshitBench V2 | 28% | #73 of 122 | Official leaderboard | |
| LMArena (Elo) | 1,314 | #215 of 333 | Official leaderboard | |
| RULER | 86.3% | #25 of 38 | Model card | |
| Humanity's Last Exam | 10.6% | #133 of 147 | Model card | |
| MultiChallenge | 38.5% | #33 of 36 | Model card | |
| SEA-HELM | 46.9 | #57 of 62 | Official leaderboard | |
| MMLU-ProX | 59.5% | #16 of 17 | Model card | |
| AA-LCR (Artificial Analysis Long Context Reasoning) | 35.9% | #30 of 31 | Model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2025 | 89.1% | #55 of 143 | Model card | |
| GPQA Diamond | 73% | #152 of 284 | Model card | |
| miniF2F | 50% | #5 of 5 | Model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveCodeBench v6 | 68.3% | #53 of 76 | Model card | |
| SciCode | 33.3% | #40 of 47 | Model card | |
| SWE-bench Verified | 38.8% | #142 of 150 | Model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LLM Chess | -200 | #93 of 124 | Official leaderboard | |
| BFCL V4 | 53.8% | #21 of 24 | Model card | |
| Terminal-Bench Hard | 8.51% | #10 of 11 | Model card | |
| tau2-bench | 49% | #66 of 69 | Model card | |
| GAIA | 0% | #57 of 57 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Agentic BAIM-LLM Evaluation (ABLE) | 24.2% | #13 of 16 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 36.7% | #59 of 71 | Official leaderboard |
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.
Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning 23 Dec 2025
Capabilities
- MMLU-Pro 78.3
- AIME 2025 89.1
- GPQA Diamond 73
- LiveCodeBench v6 68.3
- SciCode 33.3
- Humanity's Last Exam 10.6
- miniF2F 50
- Terminal-Bench Hard 8.51
- SWE-bench Verified 38.8
- tau2-bench 49
- BFCL V4 53.8
- IFBench 71.5
- MultiChallenge 38.5
- Arena-Hard v2.0 67.7
- AA-LCR (Artificial Analysis Long Context Reasoning) 35.9
- RULER 86.3
- MMLU-ProX 59.5
- WMT24++ 86.2