Back to AIBriefs
LaunchAI Models

Google introduces Gemma 4 12B, an encoder-free multimodal model

Google introduces Gemma 4 12B, an encoder-free multimodal modelGoogle Blog (General)

The 12B-parameter model runs locally on laptops with 16GB VRAM, features native audio inputs, and is released under Apache 2.0. It delivers performance nearing the larger 26B MoE model while supporting advanced reasoning and agentic workflows.

·
Jun 3, 4:00 PM
Google introduces Gemma 4 12B, an encoder-free multimodal model — AIBriefs