Multimodal
An AI system that processes more than one kind of input — text plus images, or text plus audio, or all four modes together.
A multimodal model can take multiple kinds of input in the same request. Text plus images is the most common combination: you paste a photo and ask questions about it, or drop a screenshot and ask what's wrong with the UI. Some systems extend to audio (voice input), video (short clips), and PDFs (which are technically text but often need vision-model handling for layout).
The term is used loosely. A model that can process images but only generates text is often called "multimodal" even though its output is single-mode. A truly bidirectionally-multimodal model would take images and audio and text as input and produce all three as output — that's still relatively rare in 2026.
Practical relevance: multimodal is why you can drop a screenshot of a spreadsheet into a chat and ask about it, or share a photo of a broken part and get diagnostic help. If a model was text-only, both would require you to describe what you were looking at.