Credit Consumption

The Parse and Extract APIs use complexity-aware pricing: credit consumption scales with how much content a request processes, so simpler documents consume fewer credits and denser documents consume more. Rates depend on the service tier: the standard tier consumes half the credits of the priority tier. All credit totals are rounded up to the nearest 0.1 credit. For plan and billing details and pricing for other ADE tools, see Plans & Billing.

Parse

Synchronous Parse requests run on the priority tier. Asynchronous Parse Jobs requests run on either the priority or standard tier. Rates depend on the parsing model family and service tier. Each page’s credit consumption is the sum of two components: a per-page rate and an output-character rate.

DPT-3 Pro

DPT-3 Pro is the default, highest-quality parsing model. A typical business document parsed with DPT-3 Pro uses a median of 1.5 credits on the standard tier.

Component	Priority (sync and jobs)	Standard (jobs only)
Page	1 credit per page	0.5 credits per page
Output characters	0.5 credits per 1,000 output characters	0.25 credits per 1,000 output characters

DPT-3 Verity

DPT-3 Verity is the lower-latency preview model for digitally created text documents.

Component	Priority (sync and jobs)	Standard (jobs only)
Page	0.3 credits per page	0.15 credits per page
Output characters	0.2 credits per 1,000 output characters	0.1 credits per 1,000 output characters

How Output Characters Are Counted for Parse

Output characters are all characters (Unicode code points) in the returned markdown field, including the page-break and document-ID markers.

Spreadsheets

Component	Priority (sync and jobs)	Standard (jobs only)
Sheet	1 credit per sheet	0.5 credits per sheet
Output characters	The output-character rate applies to text read from embedded images; cell text is covered by the flat per-sheet rate.	The output-character rate applies to text read from embedded images; cell text is covered by the flat per-sheet rate.
Embedded image or page (a scanned page stored as an image in the sheet)	DPT-3 Pro or DPT-3 Verity rates	DPT-3 Pro or DPT-3 Verity rates

The metadata.output_markdown_chars field reports every Markdown character in the response, including cell text, so for a workbook it is higher than the output characters that were charged.

Verify Parse Credit Consumption

The response’s metadata shows the credits consumed and every number used to calculate the total. For example, a 12-page synchronous DPT-3 Pro parse could report:

"metadata": {
  "page_count": 12,               // 12 pages × 1 credit = 12 credits
  "output_markdown_chars": 48120, // 48,120 / 1,000 × 0.5 = 24.06 credits
  "billing": {
    "service_tier": "priority",   // the rate column that applied
    "total_credits": 36.1         // 12 + 24.06 = 36.06, rounded up to 36.1
  }
}

For the full field reference, see Parse API Response.

Extract

Synchronous Extract requests run on the priority tier. Asynchronous Extract Jobs requests run on either the priority or standard tier. A typical business document uses a median of 1.5 credits on the standard tier. A request’s credit consumption is the sum of the two components below.

Component	Priority (sync and jobs)	Standard (jobs only)
Input characters	1 credit per 2,500 characters	1 credit per 5,000 characters
Output characters	1 credit per 500 characters	1 credit per 1,000 characters

How Input and Output Characters Are Counted for Extract

Input characters are the characters (Unicode code points) in the Markdown you send. Output characters are the characters (Unicode code points) in the returned extraction field, serialized as compact JSON. Whitespace between values is not counted, and each non-ASCII character counts as one character, not as an escape sequence.

Verify Extract Credit Consumption

The response’s metadata shows the credits consumed and every number used to calculate the total. For example, an extract job on the standard tier could report:

"metadata": {
  "input_markdown_chars": 40000,   // 40,000 / 5,000 = 8 credits at the standard rate
  "output_extraction_chars": 1842, // 1,842 / 1,000 = 1.842 credits at the standard rate
  "billing": {
    "service_tier": "standard",    // the rate column that applied
    "total_credits": 9.9           // 8 + 1.842 = 9.842, rounded up to 9.9
  }
}

For the full field reference, see Extract API Response.

Ground

Ground requests consume no credits. Each call is still recorded in your usage history, and the response’s metadata.billing reports "total_credits": 0.0.

Zero Data Retention

Using Zero Data Retention (ZDR) with the Parse and Extract APIs does not consume any additional credits.

Get Started

Client Libraries

Parsing

Extraction

Grounding

General

Security

Administration

Pricing & Credits

Credit Consumption

Parse

DPT-3 Pro

DPT-3 Verity

How Output Characters Are Counted for Parse

Spreadsheets

Verify Parse Credit Consumption

Extract

How Input and Output Characters Are Counted for Extract

Verify Extract Credit Consumption

Ground

Zero Data Retention

​Parse

​DPT-3 Pro

​DPT-3 Verity

​How Output Characters Are Counted for Parse

​Spreadsheets

​Verify Parse Credit Consumption

​Extract

​How Input and Output Characters Are Counted for Extract

​Verify Extract Credit Consumption

​Ground

​Zero Data Retention

Parse

DPT-3 Pro

DPT-3 Verity

How Output Characters Are Counted for Parse

Spreadsheets

Verify Parse Credit Consumption

Extract

How Input and Output Characters Are Counted for Extract

Verify Extract Credit Consumption

Ground

Zero Data Retention