PDF Inputs

PDF support will let you pass full documents — contracts, research papers, reports, invoices — directly to a model alongside your question. The model reads the document and responds in context, without you needing to pre-extract text.

Planned format#

PDF inputs will use the same messages[].content array format as image inputs. A new part type — document — will carry the file:

{
  "role": "user",
  "content": [
    {
      "type": "document",
      "document": {
        "type": "pdf",
        "source": {
          "type": "base64",
          "media_type": "application/pdf",
          "data": "<base64-encoded PDF>"
        }
      }
    },
    {
      "type": "text",
      "text": "Summarize the key obligations in this contract."
    }
  ]
}

URL-referenced PDFs (passing a public HTTPS URL instead of base64) are also planned where the underlying model supports it.

Supported models#

PDF support will be available on models that explicitly expose the documents or pdf capability. Check the catalog with ?capability=pdf once the feature is live. The model detail page will indicate whether it supports PDFs, and any page-count or file-size limits.

Use cases#

• Contract and legal document review
• Financial report analysis
• Research paper Q&A
• Invoice extraction and data structuring
• Multi-page form parsing

Pricing#

PDF inputs will be billed per token like text and images. The token count depends on the number of pages and their complexity. Token costs are displayed in DA on the model detail page.