# Vision Language Model (VLM)

> An AI that can look at a photo. Then it tells you about it in words.

**Field:** Artificial Intelligence  
**Category:** Multimodal  
**Also known as:** VLM, Vision-Language Model, Visual Language Model

## Think of it like

It is like a friend who can look at a photo with you. Then you can ask them what is in it.

## Definition

A vision language model is an AI model that can take in pictures as well as text. It answers in text. First, an image encoder turns the picture into tokens, small bits the model can read like words. The model then reads those tokens along with your question. So it can describe a photo, read a chart, or answer questions about a screen. Most big chat models today can do this.

## Example

Emma sends a photo of a broken bike chain to a chat app and asks what is wrong. The model sees that the chain has come off the back gear and tells her how to put it back.

## Why it matters

Much of the world is pictures, charts and screens, not text. A model that can see can help with all of them.

## Common mistake

Trusting it to read small text or count many things exactly. It can miss fine detail, so check numbers it reads from a picture.

## Three principles

- **Sees and reads.** Pictures and text.
- **Image to tokens.** Then read like words.
- **Answers in text.** Describe, read, explain.

## Related terms

- [Multimodal Model](https://learn.benibix.com/term/ai/multimodal-model)
- [Large Language Model (LLM)](https://learn.benibix.com/term/ai/large-language-model)
- [Image Captioning](https://learn.benibix.com/term/ai/image-captioning)
- [Visual Question Answering](https://learn.benibix.com/term/ai/visual-question-answering)
- [Token (AI)](https://learn.benibix.com/term/ai/token)
- [Vision-Language-Action Model (VLA)](https://learn.benibix.com/term/ai/vision-language-action-model)
- [Video Understanding](https://learn.benibix.com/term/ai/video-understanding)

## Told apart from

- [VLA vs VLM](https://learn.benibix.com/compare/vision-language-action-model-vs-vision-language-model): People often think a vision-language-action model is just a vision language model in a robot. Both look at pictures and read words. A vision language model answers in words, such as 'the cup is on the left'. An action model outputs moves, such as how far to turn each joint to pick the cup up. Most are built on a vision language model, then trained on clips of robots at work.
- [Multimodal Embedding vs Vision Language Model](https://learn.benibix.com/compare/multimodal-embedding-vs-vision-language-model): People think a multimodal embedding model can describe a picture. It cannot write a word. It turns a picture or some text into numbers, so you can measure how close they are. A vision language model reads the picture and writes an answer in words.

![A photo goes into an image encoder, which turns it into tokens. The tokens join the text question and go into a language model, which writes the answer.](https://learn.benibix.com/diagrams/mm-vision-language-model.svg)

---

Source: https://learn.benibix.com/term/ai/vision-language-model
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-17
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
