A 30B multimodal agentic model — answering questions, reasoning over images and video,and completing tasks, right here on ZeroGPU.