Models & pricing

Parse requests meter pages at the retail rate of the ocr_model you select. Credits deduct from your account balance as you use the API. New accounts start with $10 in free credits, about 10,000 pages on open-weight hosted models at $1 per 1,000 pages. There are no minimums or per-seat fees.

No. Each ocr_model publishes its own page rate. Open-weight hosted models like paddleocr-vl-1.6 start at $1 per 1,000 pages. Mistral, Azure, Google, and AWS Textract models list higher rates in the catalog table and on GET /models/ocr.

Some models expose optional controls, for example layout depth or table handling. GET /models/ocr lists each model's option controls and any surcharge they add to the base page price. The catalog table shows default-option retail rates.

No. Parse meters pages at the OCR model rate. Extract meters LLM tokens at the rate of whichever llm_model you pick. You can parse without extracting, or extract from a file-backed graph produced by any ocr_model.

New accounts receive $10 in free credits at signup. Credits apply to parse and extract usage until the balance is exhausted. At $1 per 1,000 pages on open-weight hosted models, that covers roughly 10,000 parsed pages before you add payment.

Yes. Change ocr_model on POST /parse, /parse/async, or /parse/batch. Endpoints, auth, jobs API, and the openparser@1 output shape stay the same. Only the per-page rate changes to match the model you route to.