A research lab developing advanced multimodal models that can understand and generate text, images, and video.