A preprint reports that DIPrune cuts visual tokens while preserving benchmark performance and reducing time to first token in multimodal AI tests.