First multimodal Gemma (vision + text). 128K token context (4B+). 140+ languages. Distilled training from larger Gemini models. Gemma-3-4B competitive with Gemma-2-27B; Gemma-3-27B comparable to Gemini 1.5 Pro.

Sizes: 270M, 1B, 4B, 12B, 27B. The 27B variant became one of the most popular open models for fine-tuning and deployment. Apache 2.0.

Model Details

Architecture DENSE
Parameters 27B
Context window 128,000

Variants

Name Parameters Notes
Gemma 3 1B 1B
Gemma 3 4B 4B
Gemma 3 12B 12B
Gemma 3 27B 27B

Paper

arXiv: 2503.19786

open-weightopen-sourcemultimodal

Related