A compact, quantized variant of Google's Gemma 4 family, running with 4-bit weights and 16-bit activations (W4A16) to reduce memory footprint while preserving much of the original model's behavior. The quantization trade-off means slightly reduced precision compared to full-precision counterparts, but it fits into tighter hardware constraints. It handles text-in, text-out tasks and is distributed in the safetensors format for straightforward loading.