Tag
We just launched Gemma 4 12B, a mid-sized multimodal model with native audio inputs, requiring only 16GB memory and released under Apache 2.0.