Model index
Google DeepMind·Gemma

Gemma 4 12B

A unified encoder-free multimodal model for laptops, filling the gap between E4B and 26B.

CurrentLanguageReasoningAgentMultimodalOpen weights
MODEL NUMBER107
Release date
2026-06-03
Organization
Google DeepMind
Family
Gemma
01

Specifications

Only published specifications are shown.

Total parameters
12B
Architecture
Dense
Context
128K
Input modalities
Text · Image · Audio · Video
Output modalities
Text
Open weights
Yes
License
Apache-2.0
Status
Current
02

Overview

Overview

A unified encoder-free multimodal model for laptops, filling the gap between E4B and 26B.

03

Changes from predecessor

Compared with Gemma 4 31B

PreviousGemma 4 31B31B · 256K ctx
CurrentGemma 4 12B12B · 128K ctx
New capabilities On-device Native multimodal
04

Architecture & capabilities

Key capabilities

On-deviceNative multimodalTool use
05

Release history

06

Related models

Select a card to view the related model.

07

Sources