Benchmark Almanac

Gemma 3n E4B

Gemma 3n E4B.

Google DeepMind Released June 26th, 2025 Family Gemma 3n The lab's model card Compare with others
37
Benchmarks with a score
0
Ranked first, out of five or more models
0
In the top three
Google DeepMind Jun 2025

Gemma 3n E4B

Open weights Small

107.2 #384 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • Audio MC 15.5% #16 /18
  • LiveCodeBench v5 25.7% #46 /47
  • MMAU-Pro 39.7% #9 /10

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
BoolQ 81.6% #14 of 28 Lab model card
ARC-Easy 81.6% #14 of 23 Lab model card
LMArena (Elo) 1,317 #214 of 335 Official leaderboard
ARC-Challenge 61.6% #36 of 49 Lab model card
TriviaQA 70.2% #30 of 40 Lab model card
SIQA 50% #18 of 24 Lab model card
PIQA 81% #28 of 37 Lab model card
HellaSwag 78.6% #41 of 53 Lab model card
WinoGrande 71.7% #38 of 47 Lab model card
Global MMLU Lite 64.5% #13 of 16 Lab model card
MMLU 64.9% #111 of 131 Lab model card
Natural Questions 20.9% #17 of 20 Lab model card
BIG-Bench Hard 52.9% #63 of 73 Lab model card
INCLUDE 57.2% #27 of 31 Lab model card
MGSM 60.7% #50 of 57 Lab model card
Wer wird Millionär? LLM Benchmark 111.0 #68 of 77 Official leaderboard
DROP 60.8% #37 of 41 Lab model card
MMLU-Pro 50.6% #154 of 167 Lab model card
MMMLU 35.6% #71 of 71 Lab model card
MMLU-ProX 19.9% #23 of 23 Lab model card
Global MMLU 60.3% #14 of 14 Lab model card

Math and science

BenchmarkScoreRankSource
HiddenMath 37.7% #9 of 11 Lab model card
DTBench 49.6% #166 of 180 Official leaderboard
Misguided Attention 22.8% #44 of 45 Official leaderboard
GPQA Diamond 23.7% #300 of 302 Lab model card
AIME 2025 11.6% #143 of 143 Lab model card

Coding

BenchmarkScoreRankSource
HumanEval 75% #62 of 101 Lab model card
MBPP 63.6% #53 of 71 Lab model card
LiveCodeBench 13.2% #52 of 54 Lab model card
LiveCodeBench v5 25.7% #46 of 47 Lab model card

Vision, audio and robotics

BenchmarkScoreRankSource
Audio MC 15.5% #16 of 18 Official leaderboard
MMAU-Pro 39.7% #9 of 10 Official leaderboard
MMAU v05.15.25 58% #6 of 6 Official leaderboard

Safety and security

BenchmarkScoreRankSource
nocot-bench (NCRI) 38.5 #178 of 185 Official leaderboard

Other

BenchmarkScoreRankSource
ACCoRD 41.2 #131 of 152 Official leaderboard
Conceptual Reasoning Index 13.3 #141 of 151 Official leaderboard
LMCA 3.16% #146 of 147 Official leaderboard

Its lab's numbers put it at 107.2, number 384 of 404

The Self-Reported Index fits one score to the 26 capability results Google DeepMind 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 105.2 to 109.2. How we compute it.

Its launch post

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

Introducing Gemma 3n: The developer guide (Google Developers Blog) and Gemma 3n model card Jun 26, 2025