A model component that processes video frames and converts them into compact numerical representations that capture the video's visual and motion content.
Performance retention over long documents and conversations
Quality of vision, audio, and image understanding (distinct from modality support)