A compact multimodal model that punches above its weight class for its size, handling both text and image inputs within a quantized 8-bit footprint. The reduced precision keeps memory usage lean, making it practical for local deployment on consumer hardware. Expect capable general reasoning and vision understanding, with the usual trade-offs that come from aggressive quantization — some loss in nuance compared to full-precision counterparts.