Image Inputs

Vision models on OpenDunes accept images directly in the messages[].content field. Instead of passing content as a plain string, pass an array of typed parts — text and image mixed in any order.

Supported models#

Any model with the vision capability accepts image inputs. Query the catalog:

GET https://opendunes.com/api/models?capability=vision

Browse vision models at /models. Well-known options include anthropic/claude-opus-4.8, anthropic/claude-sonnet-5, and google/gemini-2.5-pro.

Message format#

Replace the content string with an array of parts. Each part is an object with a type field:

• { "type": "text", "text": "..." } — a text segment
• { "type": "image_url", "image_url": { "url": "..." } } — an image

You can interleave text and images in the same content array.

Using a public URL#

Pass an HTTPS URL directly:

{
  "model": "anthropic/claude-opus-4.8",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "image_url",
          "image_url": {
            "url": "https://upload.wikimedia.org/wikipedia/commons/thumb/4/47/PNG_transparency_demonstration_1.png/240px-PNG_transparency_demonstration_1.png"
          }
        },
        {
          "type": "text",
          "text": "Describe this image in detail."
        }
      ]
    }
  ]
}

Using base64#

Encode your image as base64 and embed it as a data URI. The format is data:<mime_type>;base64,<data>. Supported MIME types: image/jpeg, image/png, image/gif, image/webp.

{
  "type": "image_url",
  "image_url": {
    "url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA..."
  }
}

Full examples#

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://opendunes.com/v1",
    api_key=os.environ["OPENDUNES_API_KEY"],
)

resp = client.chat.completions.create(
    model="anthropic/claude-opus-4.8",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {"url": "https://example.com/chart.png"},
                },
                {"type": "text", "text": "What does this chart show?"},
            ],
        }
    ],
    max_tokens=512,
)
print(resp.choices[0].message.content)

Multiple images#

You can pass multiple image parts in a single message. The model processes them all:

{
  "role": "user",
  "content": [
    { "type": "image_url", "image_url": { "url": "https://example.com/before.png" } },
    { "type": "image_url", "image_url": { "url": "https://example.com/after.png" } },
    { "type": "text", "text": "What changed between these two screenshots?" }
  ]
}

Tips#

• Resolution: Larger images consume more tokens and cost more. Resize to the minimum resolution your task requires before sending.
• Base64 vs URL: URLs are simpler but require the image to be publicly reachable. For private images, use base64.
• Model selection: Not all models support vision — check capability=vision in the catalog before choosing a model. Sending an image to a non-vision model will return an invalid_request error.
• Non-vision models: Text-only messages work the same as always — content as a plain string is still valid.

Token costs#

Image tokens are billed in DA at the same per-token rate as text. The number of image tokens depends on the model and the image dimensions. Check the model detail page at /models for the exact image token pricing.