Docs / API reference

Vision input

Images are passed as image_url, supporting both public URLs and base64.

JSON

"content": [
  { "type": "text", "text": "What's in this image?" },
  { "type": "image_url", "image_url": { "url": "https://example.com/a.png" } }
]

Limits

Up to 20MB per image and 10 images per request.
Images are converted to input tokens for billing according to each upstream's rules.
Filter by the "Vision" tag in the model plaza to see every vision-capable model.