AI models that generate images from text descriptions or prompts.
Quality of vision, audio, and image understanding (distinct from modality support)