A mid-sized multimodal model that can process both text and images, quantized to INT4 precision to reduce memory footprint while preserving much of the original capability. The compression trade-off means it runs on more modest hardware than its full-precision counterpart, though some fidelity loss is inherent to aggressive quantization. It operates in a hybrid thinking/non-thinking mode typical of the Qwen 3 family, giving it flexibility across reasoning-heavy and conversational tasks.