Google Gemma 4 (12B)
PulseAugur coverage of Google Gemma 4 (12B) — every cluster mentioning Google Gemma 4 (12B) across labs, papers, and developer communities, ranked by signal.
-
NVIDIA launches 550B open-weight Nemotron 3 Ultra for enterprise
NVIDIA has released Nemotron 3 Ultra, a 550 billion parameter open-weight model designed for enterprise applications like agents and long-context code analysis. This model is notable for being the largest permissively l…
-
Google Gemma 4 12B performance boosted by quantization techniques
A blog post compares the performance of the Google Gemma 4 12B model with and without quantization techniques, specifically MTP (Mixed Precision Training) and QAT (Quantization-Aware Training). The author provides speed…
-
Google DeepMind releases multimodal Gemma 4 12B models
Google DeepMind has released several variants of its Gemma 4 models, including the 12B parameter versions. These models are multimodal, capable of processing text, image, audio, and video inputs, with a focus on efficie…