A compact reasoning-focused model that punches above its weight class through distillation from a larger teacher model. It handles both text and image inputs, making it multimodal despite its modest 9B parameter footprint. As a distilled model, it trades some raw capability ceiling for efficiency and accessibility.