MULTIMODAL
A model that works with more than one kind of information.
WHAT DOES IT ACTUALLY MEAN?
A multimodal model can receive or produce more than one kind of information: text, images, audio, video, spatial data or actions. Some systems translate each mode into a shared internal representation; others connect specialist components.
WHY ARE WE TRACKING IT?
Conscious experience is deeply integrated across senses. Multimodality therefore matters to the investigation—but accepting several file types is not evidence of a unified inner world.
EXAMPLES IN THE FIELD.
Google’s Gemini 3.1 Pro accepts text, images, audio and video and produces text. A user can ask it to describe a video alongside its spoken audio.
RECEIPTS & REVISIONS.
Use this definition as the reading key for Model Watch. A capability claim only moves an AICNW assessment when the dated evidence meets the framework—not because a model said something dramatic on the internet.