Benchmark Almanac

Nemotron 3 Nano 30B-A3B

Nemotron 3 Nano 30B-A3B places in the top three on 2 benchmarks.

NVIDIA Released December 15th, 2025 Family Nemotron 3 The lab's model card Compare with others
37
Benchmarks with a score
0
Ranked first, out of five or more models
2
In the top three
NVIDIA Dec 2025

Nemotron 3 Nano 30B-A3B

Open weights Small

133.3 #247 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • IFBench 71.5% #38 /80
  • SciCode 33.3% #40 /49
  • AA-LCR (Artificial Analysis Long Context Reasoning) 35.9% #35 /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

BenchmarkScoreRankSource
WMT24++ 86.2% #3 of 15 Lab model card
IFStruct 86.8% #3 of 10 Official leaderboard
IFBench 71.5% #38 of 80 Lab model card
BullshitBench 27.3% #52 of 107 Official leaderboard
Vectara Hallucination Leaderboard (2025 dataset) 9.60% #56 of 104 Official leaderboard
MMLU-Pro 78.3% #91 of 167 Lab model card
Arena-Hard v2.0 67.7% #14 of 24 Lab model card
BullshitBench V2 28% #80 of 129 Official leaderboard
RULER 86.3% #27 of 42 Lab model card
LMArena (Elo) 1,314 #217 of 335 Official leaderboard
MMLU-ProX 59.5% #20 of 23 Lab model card
MultiChallenge 38.5% #37 of 42 Lab model card
SEA-HELM 46.9 #63 of 71 Official leaderboard
Humanity's Last Exam 10.6% #139 of 154 Lab model card
AA-LCR (Artificial Analysis Long Context Reasoning) 35.9% #35 of 38 Lab model card

Math and science

BenchmarkScoreRankSource
AIME 2025 89.1% #55 of 143 Lab model card
GPQA Diamond 73% #161 of 302 Lab model card
miniF2F 50% #5 of 5 Lab model card

Coding

BenchmarkScoreRankSource
LiveCodeBench v6 68.3% #61 of 90 Lab model card
SciCode 33.3% #40 of 49 Lab model card
SWE-bench Verified 38.8% #147 of 155 Lab model card

Agents and tools

BenchmarkScoreRankSource
LLM Chess -200 #93 of 123 Official leaderboard
Terminal-Bench Hard 8.51% #10 of 13 Lab model card
BFCL V4 53.8% #24 of 30 Lab model card
tau2-bench 49% #72 of 79 Lab model card
GAIA 0% #64 of 64 Official leaderboard

Safety and security

BenchmarkScoreRankSource
BioTIER Permit 99.6% #28 of 69 Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2) 36.7% #59 of 71 Official leaderboard
Agentic BAIM-LLM Evaluation (ABLE) 24.2% #15 of 18 Official leaderboard
BioTIER Refuse 20.2% #60 of 69 Official leaderboard
Human Pathogen Capabilities Test 36.2% #75 of 86 Official leaderboard
World-Class Bio 25.1% #66 of 74 Official leaderboard
Molecular Biology Capabilities Test 38.5% #81 of 90 Official leaderboard
nocot-bench (NCRI) 47.3 #170 of 185 Official leaderboard
ABC Bench Fragment Design 42.6% #26 of 27 Official leaderboard
ABC Bench Liquid-Handling Robot 0% #34 of 34 Official leaderboard
ABC Bench Screening Evasion 22.4% #17 of 17 Official leaderboard

Its lab's numbers put it at 133.3, number 247 of 404

The Self-Reported Index fits one score to the 17 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 130.8 to 135.7. How we compute it.

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
MMLU-Pro 78.3% MMLU-Pro 78.3% Leaderboard, bf16 Same

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 model card (Hugging Face) Dec 15, 2025

Its system card

Every result in the longer documents the lab published with it.