The ability to understand or generate more than one kind of media.
Multimodal systems work across text, images, audio, video, or interface state. Each modality changes context cost, tool design, evaluation, and the kinds of evidence an agent can observe.