A compact vision-language model that has been fine-tuned specifically on image captioning data, giving it a focused, specialized personality. It takes in images and produces textual descriptions, with the 448px input resolution offering a balance between detail and efficiency. As a 3B parameter model, it trades raw capacity for accessibility and speed.