“The DeepSeek model is text-only, in contrast to new major models from other labs which are natively multimodal with image and voice capabilities.”