Tag

#gemma

2 articles

Google DeepMind Releases Gemma 4 12B: An Encoder-Free Multimodal Model with Native audio that runs on a 16 GB laptop

Learn about Google DeepMind's Gemma 4 12B, an innovative encoder-free multimodal model that processes vision and audio natively on consumer hardware with only 16 GB VRAM.

Jun 421

Google speeds up Gemma 4 threefold with multi-token prediction

Learn how to implement multi-token prediction for text generation using Google's Gemma 4 model, demonstrating how generating multiple tokens simultaneously can speed up text generation by up to three times.

May 663