← Back to the A-Z glossary
AI glossary · P1

Vision model

A vision model understands or describes images.

Modality 4 min read Updated 2026-05-22
— Definition

Vision model, in plain words

A vision model understands or describes images.

Present AI vision as OCR boosted with reasoning, not as an object detector.

A concrete example

I send Claude a screenshot of a dashboard and ask 'what's wrong with this chart'. It answers with real analysis.

Why it matters

AI vision replaces 80% of classic OCR cases and unlocks reading visual docs (invoices, screens, diagrams).

You run into it when analyzing screenshots, photos, charts, tables, and scanned documents.

Don't mix it up with

Multimodal: A multimodal model accepts several types of input, like text, images, audio, or video.

OCR: OCR turns the text inside an image or scan into text a machine can read.

Common mistakes

  • Asking it to count elements precisely (a known weak spot).
  • Forgetting to resize images (cost and slowness).
  • Mixing up vision (reading) and image generation (creating).

Quick checklist

  • First I check whether the word names a concept, a tool, a risk, or a metric.
  • I tie it to a concrete case: I send Claude a screenshot of a dashboard and ask 'what's wrong with this chart'. It answers with real analysis.
  • I keep the main trap in mind: Asking it to count elements precisely (a known weak spot).

Quick questions

What is Vision model in AI?

A vision model understands or describes images.

Where will I run into Vision model?

You run into it when analyzing screenshots, photos, charts, tables, and scanned documents.

Which word should I read next?

Start with Multimodal, OCR, AI model.

Want to keep going in order?

Head back to the full glossary, search a word, then open only the pages that deserve more than a short definition.

Open the AI glossary