[whatmodel.sh](../../README.md) / Gemma 3n E4B # Gemma 3n E4B Google · released 2025-06 · 7.85B params (4B active) · gemma **score 49.1** ±1.3 · #389 of 732 · from 12 benchmark results across 4 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at _{● marks Gemma 3n E4B among every model on that task or test, worst to best.}

task            vs the others          rank  counts  tests
coding          █··········●        #1 of 2     21%      3
chat & writing  ▁▁▁▃▆▄▆●▆▇▇▃   #207 of 402*     19%      1
math            █··········●       #1 of 2*     25%      2

coding · #1 of 2 · 21% of its score

test                                result  vs the others in its run    difficulty      counts  why
HumanEval (maker)                    75.0%  ▁▁▁▁▂▂▆▆●▅▂▁ beats 158/205  easy · avg 78%     11%  sharp · small run
MBPP (maker)                         63.6%  ▁▁▁▄▅█▆●▄▁▂▁ beats 59/86    easy · avg 68%    5.7%  at its level · small run
LiveCodeBench v5 (window unspecif…   25.7%  ▃●·····▃█··▅ beats 1/7      hard · avg 32%    4.2%  at its level · small run

chat & writing · #207 of 402 · 19% of its score

Beats [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Granite 4.2 8B](granite-4-2-8b.md), [Mistral Small 4 (119B-A6.5B)](mistral-small-4.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md) here, though they rank lower overall.

test            result  vs the others in its run    difficulty  counts  why
LMArena (text)  1317.5  ▁▁▁▂▃▃▃●▆▇█▃ beats 148/385  avg 1260       19%  blunt

math · #1 of 2 · 25% of its score

test               result  vs the others in its run  difficulty      counts  why
AIME 2025 (maker)   11.6%  ▁·▁▁●▂▅▆█▇▅▁ beats 6/133  hard · avg 17%     13%  sharp · small run
MGSM (maker)        60.7%  ▂·▁▁▁▁▁▂▅█●▂ beats 35/42  easy · avg 67%     12%  sharp · at its level · small run

other tests (8)

test                                result  vs the others in its run    difficulty        counts  why
Open Persian LLM Leaderboard v2 a…   55.0%  ▂▄▂▃█●▃▆▃▅▂▂ beats 40/81    medium · avg 57%    8.8%  blunt · at its level
MMLU (maker)                         64.9%  ▁▁▁▁▁▂▂▅●█▆▄ beats 105/214  medium · avg 70%    8.6%  at its level · small run
MMLU-Pro (maker)                     50.6%  ▁▁·▁▁▁▃●▄▅█▁ beats 51/188   medium · avg 53%    7.0%  at its level · small run
Global-MMLU (maker)                  60.3%  ▃●▃·▃···▅·▅█ beats 1/11     medium · avg 63%    3.9%  at its level · small run
EuroEval HellaSwag (English), tes…   71.3%  ▃·▁▁▂▁▂▅●▇▄▁ beats 54/100   medium · avg 67%    3.7%  sharp · at its level · noisy
EuroEval Life in the UK (English …   80.2%  ▁▁▁▁▁▁▁▂▄●▄▃ beats 66/146   easy · avg 82%      2.4%  at its level · noisy
Codegolf v2.2 (maker)                16.8%  █··········● beats 1/1                             —  not counted yet: no other model in the …
GPQA Diamond (Google RelaxedAccur…   23.7%  ······●····· beats 0/1                             —  not counted yet: no other model in the …

## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md)

source                                    results  share of the score
Google DeepMind Gemma 3n E2B IT model c…        8   65% ▇▇▇▇▇▇▇▇
LMArena leaderboard dataset                     1   19% ▇▇
Open Persian LLM Leaderboard v2                 1    9% ▇
EuroEval leaderboards                           2    6% ▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 49.1

_{Licences: Google DeepMind Gemma 3n E2B IT model card ([CC-BY-4.0](../../licenses/cards-final-NOTICE.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Open Persian LLM Leaderboard v2 ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE))}

Not counted (2)

source                  test                            score  why
Google DeepMind Gemma…  Codegolf v2.2                   16.8%  too few models have it yet
Google DeepMind Gemma…  GPQA Diamond (Google RelaxedA…  23.7%  too few models have it yet

## How to run it ### llama.cpp (recommended) 1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md) ```sh # macOS / Linux curl -LsSf https://llama.app/install.sh | sh # Windows irm https://llama.app/install.ps1 | iex ``` 2. Run Gemma 3n E4B (Q4_K_M, needs ~4.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/gemma-3n-E4B-it-GGUF/tree/90fa8b0e431faeae50c305828bc260d6f71720e1)) ```sh llama cli -hf unsloth/gemma-3n-E4B-it-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf unsloth/gemma-3n-E4B-it-GGUF:Q4_K_M ``` Other versions: Q8_0 (~7.4 GB) `llama cli -hf unsloth/gemma-3n-E4B-it-GGUF:Q8_0` ### LM Studio 1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download) 2. Run Gemma 3n E4B (Q4_K_M, needs ~4.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/gemma-3n-E4B-it-GGUF/tree/90fa8b0e431faeae50c305828bc260d6f71720e1)) ```sh lms get https://huggingface.co/unsloth/gemma-3n-E4B-it-GGUF@Q4_K_M -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: Q8_0 (~7.4 GB) `lms get https://huggingface.co/unsloth/gemma-3n-E4B-it-GGUF@Q8_0 -y` · 4bit (~5.8 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/gemma-3n-E4B-it-MLX-4bit -y` · 8bit (~9.3 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/gemma-3n-E4B-it-MLX-8bit -y` ### Ollama 1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download) ```sh # Linux curl -fsSL https://ollama.com/install.sh | sh ``` 2. Run Gemma 3n E4B (Q4_K_M, needs ~5.21 GB, from [ollama.com](https://ollama.com/library/gemma3n/tags)) ```sh ollama run gemma3n:e4b ``` Other versions: Q8_0 (~9.92 GB) `ollama run gemma3n:e4b-it-q8_0` ### MLX (Mac) 1. Install MLX (Mac). Apple silicon Macs only, text only. ```sh # with uv uv tool install mlx-lm # or pip install mlx-lm ``` 2. Run Gemma 3n E4B (4bit, needs ~5.8 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/gemma-3n-E4B-it-MLX-4bit/tree/f6432cfcd80f2b4b0940ba96da76f6e6264f762c)) ```sh mlx_lm.chat --model lmstudio-community/gemma-3n-E4B-it-MLX-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model lmstudio-community/gemma-3n-E4B-it-MLX-4bit ``` Other versions: 8bit (~9.3 GB, Mac only) `mlx_lm.chat --model lmstudio-community/gemma-3n-E4B-it-MLX-8bit` Model repository and documentation: [huggingface.co/google](https://huggingface.co/google/gemma-3n-E4B-it) [interactive version](https://whatmodel.sh/model/gemma-3n-e4b) · [JSON](https://whatmodel.sh/api?format=json&all=1)