A compact multimodal model that punches at a small parameter count while handling both text and image inputs. At 3B parameters, it trades raw reasoning depth for efficiency and accessibility, making it a lightweight option for vision-language tasks. Details about its specific strengths and limitations in practice are limited.