Tag: Kaggle

Google unveiled Gemma 4 12B, a 12‑billion‑parameter large language model that delivers near‑parity with its 26‑billion‑parameter sibling while fitting on a laptop with 16 GB of RAM. The model introduces Multi‑Token Prediction (MTP) drafters for faster, more efficient token generation and a streamlined vision embedding that removes bulky encoders. Audio inputs bypass encoding altogether, letting raw signals feed directly into the model. Weights, just under 18 GB, are downloadable from Kaggle and Hugging Face, and the model can be accessed through LM Studio, Google AI Edge Gallery and similar tools. Leggi di più