Google Releases Open-Source Gemma412B Model: Focuses on Encoder-Free Multimodal with 16GB Memory Notebook for Local Execution

Google releases new open-source large model Gemma412B with a unified encoder-free architecture, enabling on-device full-modal AI. It directly processes text, images, audio, and video in a single Transformer backbone, eliminating memory and latency issues from external encoders.....