A compact multimodal model that handles both text and image inputs, producing text outputs. At 2 billion parameters, it operates within tight resource constraints, making it accessible for deployment on modest hardware. Its small footprint means capability trade-offs compared to larger vision-language models are expected.