Today we’re introducing Gemma 4 12B — our latest open model that brings advanced agentic reasoning, vision and audio…

4 tweets1 min read♥ 9.3Koriginal

Summary

Google introduces Gemma 4 12B, an open-source AI model that runs locally on 16GB VRAM with advanced reasoning, vision, and audio capabilities. The model uses a novel unified architecture that eliminates separate multimodal encoders, reducing memory usage and latency while delivering performance comparable to larger models.

Summarized by ThreadOut AI from the full thread. May miss nuance — read the thread below.

#1
Today we’re introducing Gemma 4 12B — our latest open model that brings advanced agentic reasoning, vision and audio directly to your laptop. It delivers performance nearing our larger Gemma models with a much smaller total memory footprint, while being small enough to run locally with just 16GB of VRAM. It’s open and accessible for everyone to use under a permissive Apache 2.0 license. This is all made possible by our new, unified architecture that removes separate multimodal encoders. Here’s how we did it 🧵
#2
Gemma 4 12B delivers great performance with a small memory footprint and a novel architecture.
#3
Our new unified architecture allows Gemma 4 12B to process multimodal inputs natively. Here's how ⬇️ Traditional models rely on separate encoders for images and audio. This adds latency and increases memory usage. So we streamlined this: 👁️ Vision: We took a novel approach to replace the encoder with a lightweight embedding module, letting the LLM backbone take over visual processing. 🎙️ Audio: We removed the encoder entirely, projecting the raw audio signal directly into the same space as text tokens.
#4
Like the rest of the Gemma 4 family, Gemma 4 12B is released under a permissive Apache 2.0 license for anyone to use. Download the model weights today on @huggingface or @kaggle. Learn more and see how to start building with Gemma 4 12B ↓ goo.gle/49FoQin