The Decoder / One term

Can Claude generate images?

No. Claude reads images but does not make them; Anthropic's own documentation says it cannot generate, edit or create images. What it can do is write a picture as code, an SVG or a chart, which your browser then draws.

By Mstimaj Numbers checked 6 November 2026 Updated 6 November 2026 Level Start Here Tool Claude

Sign in to save your place and mark this done. Free, no card.

A dark screen showing a chat reply made of code on the left and the clean vector drawing it produces on the right

No. Claude can look at a photo, a screenshot or a diagram you give it and tell you what is there, but it has no image generator. Ask it for a picture and it will either explain that, or write one as code: an SVG, a chart, a diagram inside an artifact. Those are real images, drawn by your browser from Claude's instructions, and they never go blurry. For a photograph-style picture you need ChatGPT or Gemini.

Key facts

  • 0images Claude can generate, by Anthropic's own documentation
  • 600images or PDF pages Claude can read in a single request on its current models
  • 2ways to get a picture out of it anyway: an SVG written as code, or a chart in an artifact
  • $20a month for Pro, the same price as the ChatGPT and Gemini plans that do generate images
  • platform.claude.com, Vision documentation, FAQ 'Can Claude generate or edit images?' and the context-window page for the 600-image limit; claude.com/pricing; all read 6 November 2026

What does Claude do when you ask for a picture?

Anthropic’s documentation answers this in one line: Claude is an image understanding model only. It can interpret and analyze images, but it cannot generate, produce, edit, manipulate, or create images. In practice, ask for a logo or an illustration and one of two things happens. Either Claude says it cannot draw and offers to describe the image or write a prompt for a tool that can, or it writes the picture as code. The second is more useful than it sounds. An SVG is a drawing written as instructions, and when the code lands in your browser the browser draws it, sharp at any size.

Which AI does generate images, then?

ChatGPT and Gemini both make pictures from a description, on their free plans with a cap and on their paid plans with more room. Gemini’s image model is the one Google calls Nano Banana. So the usual pattern for someone who likes Claude for writing and thinking is to keep using it for that and open one of the other two when a picture is needed. The three tools cost the same $20 a month at the ordinary paid tier, which is why which one you pay for is its own question, linked below.

The three main chat tools and pictures, at the $20 tier
Tool Makes images Reads images Draws as code (SVG, charts)
Claude No Yes Yes
ChatGPT Yes Yes Yes
Gemini Yes, Nano Banana Yes Yes

Can Claude make diagrams or charts?

Yes, and this is the half of Claude that gets overlooked. Charts, flow diagrams, simple illustrations, icons and page layouts are all things Claude can write as SVG or as a small web page, and in the Claude app they open in an artifact you can look at and download. A bar chart of your numbers, a diagram of how something connects, a title card: these come out clean and editable, because they are text underneath. What this route cannot do is a photograph or a painting. If the picture needs light, texture or a face, it needs a generator.

Can Claude read the pictures I give it?

That is the half it does well. Drop in a screenshot of an error, a photo of a form, a chart from a report, a whiteboard, and Claude reads it. On the current models a single request can carry up to 600 images or PDF pages. It will not identify people in photos, and Anthropic says it cannot reliably tell whether an image was made by AI, so do not use it as a detector. For everything else, reading pictures is a normal part of a conversation.

Images Claude can read in one request IMAGES CLAUDE CAN READ IN ONE REQUEST Current models (1M context) 600 images Older models (200k context) 100 images
Reading, not making. The limit depends on the model's context window.

Ask what did not make sense

Comments go through our own form and are read before they appear. Links are not allowed; describe it in words.

Comments are read before they appear.

The newsletter

Get the next term.

One email when something goes up. The argument, the film and the term, nothing else. One click to leave.

Have a question now? Ask Mstimaj, she knows the archive.