Vision

Send images inline to vision models.

Models with vision accept images inline.

json

{"role": "user", "content": [
  {"type": "text", "text": "What is in this picture?"},
  {"type": "image_url", "image_url": {"url": "data:image/png;base64,iVBORw0KGgo..."}}
]}

Anthropic format: {"type": "image", "source": {"type": "base64", "media_type": "image/png", "data": "…"}}. Remote image URLs are accepted only on models that fetch them upstream; otherwise you get 400 unsupported_parameter. Requests are limited to 32 MB.