Vision

Vision-capable chat models can look at images as well as read text. Use them to describe photos, read diagrams and charts, pull text out of screenshots or scanned documents, or answer questions about a picture.

Send an image

Images travel as image_url content parts inside a normal chat message, exactly as in the OpenAI API. Mix as much text and as many images as the model allows in a single message.

from pendra import Pendra

client = Pendra()

response = client.chat.completions.create(
    model="qwen3-vl:8b",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What's in this image?"},
            {"type": "image_url",
             "image_url": {"url": "data:image/png;base64,iVBORw0..."}},
        ],
    }],
)
print(response.choices[0].message.content)

import Pendra from 'pendra';

const client = new Pendra();

const response = await client.chat.completions.create({
  model: 'qwen3-vl:8b',
  messages: [{
    role: 'user',
    content: [
      { type: 'text', text: "What's in this image?" },
      { type: 'image_url', image_url: { url: 'data:image/png;base64,iVBORw0...' } },
    ],
  }],
});
console.log(response.choices[0].message.content);

Response

A vivid impressionist oil painting of rolling hills at sunset, built up in thick palette-knife strokes of orange, yellow, and blue. A winding path leads through fields towards the hills, with a few small figures walking along it and autumn trees and a wooden fence in the foreground.

How images are accepted

• Base64 data URIs (data:image/png;base64,…) work out of the box — the most common and most private option, since the image never leaves your request.
• Remote http(s) URLs are opt-in on self-hosted workers (the worker would otherwise fetch arbitrary URLs from inside your network). When enabled, fetches are restricted to public hosts and are size- and time-bounded.

A model is vision-capable when it advertises the vision capability in the model list — installing a vision model from the catalogue sets this up for you. Sending an image to a model that can't see it returns a clear error rather than a blind guess.

Big vision models on multi-GPU workers

A vision model reads images with a component that runs on a single GPU, so image input always uses one card. On a worker with several GPUs, a vision model too big to fit one card still runs — its text side is spread across the cards, so you can use it for chat, coding, and tool use — but image input is turned off for it there. To use its image input, run the model on a worker where the whole model fits a single GPU. This is automatic; there's nothing to configure.

If a model shows Vision crossed out

On the Workers page, a model whose Vision chip is dimmed and crossed out can't accept images on that worker. Hover it to see why. One reason is the multi-GPU case above — the model is too big for one card and is running text-only across the cards. Another is that the worker doesn't have the model's image projector, the extra file a model needs to see; that happens to models installed before the projector became available for them, and reinstalling the model on that worker brings the projector down with it.

The third is that the worker can't run that particular model's image projector. A few models ship a kind of projector Pendra doesn't support yet, so rather than risk it the worker serves the model as text only. This one is usually specific to one size of a model, so another size of the same model may well accept images — and a later worker update restores it when support lands.

In every case the model still answers text normally, and the model list stops advertising vision for it — so a client checking capabilities before sending an image gets the right answer instead of an error.