multimodal
A model that takes more than text as input. In practice this means images: screenshots, photos, scanned PDFs, encoded into the same token stream as the prompt. Some also handle audio. Output is still usually text.
An image is converted into a few hundred to a few thousand tokens, so a screenshot can cost more of the context window than the question about it.
For local work this is the difference between a chat toy and something useful. Vision is what lets a model read the scanned document, the receipt, or the broken interface you just screenshotted, without a separate OCR pipeline in front of it.