A multimodal model that handles both text and image inputs, producing text outputs. It operates under an open-weight Apache 2.0 license, making its weights freely accessible for inspection and deployment. As a community-published quantized variant of Qwen 3, it trades some numerical precision for reduced resource requirements.