MODEL LINEAGE
中文
Model index
Google DeepMind·Gemma

Gemma 3n E4B

Used MatFormer and Per-Layer Embeddings to bring multimodal understanding to mobile devices.

LegacyLanguageMultimodalOpen weights
MODEL NUMBER167
First public date
2025-06-26
Organization
Google DeepMind
Family
Gemma
01

Specifications

Only published specifications are shown.

Total parameters
8B↗
Active parameters
4B↗
Architecture
Hybrid↗
Parameter scope
E4B denotes effective memory / deployment footprint, not ordinary MoE active parameters.↗
Context
32,000 tokens
Input modalities
Text · Image · Audio · Video
Output modalities
Text
Open weights
Yes
License
Gemma Terms
Status
Legacy
02

Overview

Release dates

— Gemma 3n fully released its edge multimodal models. ↗

03

Architecture & capabilities

Model capabilities

On-deviceNative audioVision
04

Release history

Open weights

Gemma 3n fully released its edge multimodal models.

View source
05

Related models

Select a card to view the related model.

Variant ofGemma 3 27BView details →
06

Sources

01
Introducing Gemma 3nGoogle Developers · Published 2025-06-26
Official · Verified02
Gemma 3n E4B model cardGoogle · Checked 2026-09-07
Official · Verified
Suggest a correction

Copy this template, add the field, proposed value, and primary source, and share it with the archive maintainer.