# Gemma 4 12B

> **TL;DR.** Gemma 4 12B is Google DeepMind's 12-billion-parameter open-weights multimodal model, distinguished by an encoder-free architecture that processes text, images, audio, and video through a single decoder-only transformer with no separate vision or audio encoder modules.

- **Category:** AI / Open-Source Models / Multimodal
- **Stage:** nascent
- **Age:** 2 days
- **Origin date:** 2026-06-03
- **First detected:** 2026-06-04
- **Canonical URL:** https://earlyterms.com/term/gemma-4-12b
- **Sources:** 8 primary URLs

## Definition

Gemma 4 12B is Google DeepMind's 12-billion-parameter open-weights multimodal model, distinguished by an encoder-free architecture that processes text, images, audio, and video through a single decoder-only transformer with no separate vision or audio encoder modules.

Released [June 3, 2026](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12b/) under the Apache 2.0 license, Gemma 4 12B is the first mid-sized Gemma model with native audio support, targets consumer laptops with 16 GB of RAM, and outperforms Gemma 3 27B on MMLU Pro (77.2% vs 67.6%) despite half the parameter count.

## Analogy

Think of it as a Swiss Army knife that swapped out the blades for a single fused multi-tool.

## Why it's emerging now

Google DeepMind released Gemma 4 12B on June 3, 2026 — a dense multimodal model that processes text, images, audio, and video through a single encoder-free transformer, fits in 16 GB of consumer RAM, and outperforms Gemma 3 27B on MMLU Pro. It is the first mid-sized open model with native audio and a 256K context window targeting laptop deployment.

## Related terms

- *parent:* Gemma 4
- *alias:* gemma-4
- *related:* MLX
- *related:* lm-studio
- *competitor:* Qwen3
- *competitor:* qwen3
- *related:* mtp
- *competitor:* Llama 4
- *parent:* encoder-free multimodal
- *parent:* local LLM
- *related:* Gemma 3
- *related:* agentic-ai

## Sources

1. [Google Blog — Introducing Gemma 4 12B](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12b/)
2. [Google Developers Blog — Gemma 4 12B Developer Guide](https://developers.googleblog.com/gemma-4-12b-the-developer-guide/)
3. [Hugging Face Blog — Welcome Gemma 4](https://huggingface.co/blog/gemma4)
4. [Hugging Face — google/gemma-4-12b-it model card](https://huggingface.co/google/gemma-4-12b-it)
5. [Hacker News — Gemma 4 12B launch thread (973 points)](https://news.ycombinator.com/item?id=48385906)
6. [The Decoder — Gemma 4 12B squeezes multimodal AI onto a laptop](https://the-decoder.com/google-deepminds-gemma-4-12b-squeezes-multimodal-ai-onto-a-laptop-with-just-16-gb-of-ram/)
7. [VentureBeat — Gemma 4 12B enterprise analysis](https://venturebeat.com/technology/googles-new-open-source-gemma-4-12b-analyzes-audio-video-and-runs-entirely-locally-on-a-typical-16gb-enterprise-laptop)
8. [Google AI — Gemma releases changelog](https://ai.google.dev/gemma/docs/releases)

---
_Generated by EarlyTerms · https://earlyterms.com/term/gemma-4-12b_