Gemma 3n E2B

Name: Gemma 3n E2B
Author: Google

Google · Gemma 3n · released 2025-06-26 · data checked 2026-11-02

Gemma 3n E2B is a 5.4B-parameter dense model from Google (United States), released 2025-06-26 under the Gemma license. It scores 14.3 on the Small LM Index, #26 of 26 small models. It needs about 4 GB of memory at 4-bit with an 8K-token context.

Smaller Gemma 3n: runs in about 2B parameters’ worth of memory thanks to per-layer embeddings (MatFormer).

Small LM Index

14.3

#26 of 26 scored small models

Likely 12.4–16.5 · from 4 of 6 core benchmarks

Parameters

5.4B

2B effective

Price per 1M tokens

—

No hosted API on OpenRouter; self-host

Context window

32K

tokens

Benchmarks

Each bar shows this model's score. The tick marks the best small model on that benchmark. Sources: 6 reported by Google. How we label sources.

MMLU-Pro Lab-reported Knowledge · Instruct model, 0-shot (all scores)40.5

Class best: Qwen3.5 9B 82.5

GPQA Diamond Lab-reported Knowledge · GPQA Diamond24.8

Class best: Qwen3.5 9B 81.7

AIME Lab-reported Math · AIME 20256.7

Class best: Apriel 1.6 15B Thinker 88.0

LiveCodeBench Lab-reported Coding · LiveCodeBench v518.6

Class best: Granite 4.2 8B 73.2

MMLU Lab-reported Knowledge60.1

Class best: OLMo 3 7B Think 77.8

Multilingual MMLU Lab-reported Multilingual · Global-MMLU55.1

Class best: Gemma 3n E4B 60.3

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

Humanity's Last Exam≈ 1

SWE-bench Verified≈ 3

Specifications

Lab: Google (US · United States)
Architecture: Dense transformer
Total parameters: 5.4B
Active parameters: All (dense)
Size class: Small
Context window: 32K tokens
Input modalities: text, image, audio, video
Reasoning: Instruct (no thinking)
License: Gemma
Openness: Open weights
Memory to run: ~4 GB at 4-bit, 8K context (more sizes)
Hugging Face: 38.3K downloads (30 days) · 334 likes

Weights on Hugging Face ↗

Compare with

Gemma 3n E2B vs Gemma 4 E2B (5.1B, index 31.9)
Gemma 3n E2B vs Gemma 3 4B (4.3B, index 17.5)
Gemma 3n E2B vs Nanbeige4.2 3B (4.2B, index 70.4)
Gemma 3n E2B vs Spark-X2.6 4B (4.1B, index 67.5)

Sources

https://huggingface.co/google/gemma-3n-E2B-it

Something wrong here? Report it.

Run it locally

Uses unsloth/gemma-3n-E2B-it-GGUF (Q4_K_M). Swap the tag for another size: F16, IQ4_NL, IQ4_XS, Q2_K, Q2_K_L, Q3_K_M, Q3_K_S, Q4_0….

Ollama

ollama run hf.co/unsloth/gemma-3n-E2B-it-GGUF:Q4_K_M

llama.cpp

llama-server -hf unsloth/gemma-3n-E2B-it-GGUF:Q4_K_M

Then open http://localhost:8080 for a chat window.

MLX (Mac)

pip install -U mlx-lm
mlx_lm.chat --model mlx-community/gemma-3n-E2B-it-lm-4bit

Text only; for images use mlx-vlm.

vLLM (GPU server)

vllm serve google/gemma-3n-E2B-it

Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization	4K context	32K context
16-bit F16	12 GB	12 GB
8-bit Q8_0	6.4 GB	6.7 GB
4-bit Q4_K_M	4.0 GB	4.2 GB

Check it against your GPU, Mac or phone →