whatmodel.sh / Gemma 3n E4B

Gemma 3n E4B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Gemma 3n E4B is an open-weights language model from Google, released June 2025. Its published size is 7.85B parameters (4B active per token) and a 32k-token context window. Licence: gemma.

47.7±1.4#383 of 719 LLMsfrom 12 benchmark results across 4 sources

$ how it compares

The LLMs ranked either side of Gemma 3n E4B, out of 828 with a score, then more from Google.

#modelscorewhereRAMreleased
458Claude 3 Haiku Anthropic48±0comparison only2024-03
459GPT-4o mini Audio Preview (2024-12-17) OpenAI48±2comparison only2024-12
460Gemma 3n E4B Google48±1local~6.2GB2025-06
461Hermes 2 Theta Llama-3 70B Nous Research48±2local~47GB2024-06
462Granite 4.0 H Tiny (7B-A1B) IBM48±1local~5.6GB2025-09
more from Google
9Gemini 4 Argon (limited preview) Google82±1limited preview2026-09
23Gemini 3.8 Flash Google77±1api $1.50/M2026-09
32Gemini 3.7 Flash Google76±1api $1.50/M2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

coding█··········█ #1 of 2 · 24% of its score · 3 tests
testresultvs the others in its rundifficultycountswhy
HumanEval lab75.0% ✎▁▁▁▁▁▃▅▅█▃▂▁ beats 143/190easy · avg 81%▇▇▇▇▇ 12%sharp · small run
MBPP lab63.6% ✎▁▁▁▄▅█▄▅▃▁▂▁ beats 55/81easy · avg 71%▇▇▇ 6.3%at its level · small run
LiveCodeBench v5 (window unspecified) lab25.7% ✎▃▃·····▃█··▅ beats 1/7medium · avg 35%▇▇ 5.0%at its level · small run
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #202 of 393 * · 18% of its score · 1 test

Beats Granite 4.2 8B, Qwen3-VL 30B-A3B Thinking, Mistral Small 4 (119B-A6.5B) here, though they rank higher overall. Trails LFM2.5 1.2B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1317.4 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 147/379avg 1268▇▇▇▇▇▇▇ 18%blunt
math█··········█ #1 of 2 * · 22% of its score · 2 tests
testresultvs the others in its rundifficultycountswhy
AIME 2025 lab11.6% ✎▁·▁▁▃▂▅▆█▇▅▁ beats 6/133hard · avg 22%▇▇▇▇▇▇ 14%sharp · small run
MGSM lab60.7% ✎▃·▁·▁··▂▄█▄▄ beats 19/26easy · avg 66%▇▇▇ 8.5%at its level · small run
other tests8 not in a task · 34% of its score
testresultvs the others in its rundifficultycountswhy
MMLU lab64.9% ✎▁▁▁▁▁▂▃▅█▃▆▃ beats 85/183medium · avg 73%▇▇▇ 8.7%at its level · small run
Open Persian LLM Leaderboard v2 average (21 tasks)55.0% ✎▂▄▂▃█▄▃▆▃▅▂▂ beats 40/81medium · avg 58%▇▇▇ 8.0%blunt · at its level
MMLU-Pro lab50.6% ✎▁▁·▁▁▁▃▃▄▅█▁ beats 51/188medium · avg 56%▇▇▇ 7.7%at its level · small run
Global-MMLU lab60.3% ✎▃▅▃·▃···▅·▅█ beats 1/11medium · avg 65%▇▇ 4.5%at its level · small run
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy71.3% ✎▃·▁▁▂▁▂▅█▇▄▁ beats 54/100medium · avg 68%▇ 3.3%at its level · noisy
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy80.2% ✎▁▁▁▁▁▁▁▂▄█▄▃ beats 66/146easy · avg 83%▇ 2.3%at its level · noisy
Codegolf v2.2 lab16.8% ✎█··········█ beats 1/1—not counted yet: no other model in the same run to compare it with
GPQA Diamond (Google RelaxedAccuracy/accuracy) lab23.7% ✎······█····· beats 0/1—not counted yet: no other model in the same run to compare it with

$ where to run it

llama.cpp · recommended

version

  1. Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
    macOS / Linux
    curl -LsSf https://llama.app/install.sh | sh
    Windows
    irm https://llama.app/install.ps1 | iex

    other ways to install ↗

  2. Run Gemma 3n E4B · needs ~4.5 GB · from huggingface.co/unsloth
    chat in the terminal
    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)
    llama serve -hf unsloth/gemma-3n-E4B-it-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Gemma 3n E4B · needs ~4.5 GB · from huggingface.co/unsloth

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/unsloth/gemma-3n-E4B-it-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Gemma 3n E4B · needs ~5.21 GB · from ollama.com
    chat in the terminal
    ollama run gemma3n:e4b

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Gemma 3n E4B · needs ~5.8 GB, Mac only · from huggingface.co/lmstudio-community
    chat in the terminal
    mlx_lm.chat --model lmstudio-community/gemma-3n-E4B-it-MLX-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model lmstudio-community/gemma-3n-E4B-it-MLX-4bit

Official weights and documentation: huggingface.co/google ↗

$ how the score was worked out · Google DeepMind Gemma 3n E2B… 67%, LMArena leaderboard dataset 18%, Open Persian LLM Leaderboard… 8% and 1 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The 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's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Google DeepMind Gemma 3n E2B IT model card · lab-reported8▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 67%CC-BY-4.0
LMArena leaderboard dataset1▇▇▇▇▇▇ 18%CC BY 4.0
Open Persian LLM Leaderboard v2 (Part DP AI and AUT NLP Lab)1▇▇ 8%Apache-2.0
EuroEval leaderboards (results collection), English generative leaderboard2▇▇ 6%MIT
prior · pull toward the middle▇ 1%
weighted average= 47.7

not counted (2)

Google DeepMind Gemma 3n E2B IT model cardCodegolf v2.2 [correct]16.8%fewer than three models have this test yet, or this source ran it for this model alone
Google DeepMind Gemma 3n E2B IT model cardGPQA Diamond (Google RelaxedAccuracy/accuracy) [correct]23.7%fewer than three models have this test yet, or this source ran it for this model alone

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json