A 2B parameter multimodal embedding model from Qwen that processes both text and images for retrieval tasks.