Cheapest OCR API, compared per 1,000 pages
20 OCR models behind one request, cheapest first. The lowest paid price is $1.00 per 1,000 pages with DeepSeek-OCR.
Open-source models
Open-weight engines, hosted for you.
| Model | 1,000 pages | 100,000 pages | Output | Status |
|---|---|---|---|---|
Tesseract tesseract/tesseract | Free | Free | Markdown and JSON | Available |
DeepSeek-OCR deepseek/deepseek-ocr | $1.00 | $100.00 | Markdown and JSON with positions | Available |
Jina OCR jina/jina-ocr | $1.00 | $100.00 | Markdown and JSON | Available |
PaddleOCR-VL paddle/paddleocr-vl | $2.00 | $200.00 | Markdown and JSON with positions | Available |
GLM-OCR zai/glm-ocr | $2.00 | $200.00 | Markdown and JSON with positions | Available |
Unlimited OCR baidu/unlimited-ocr | $2.00 | $200.00 | Markdown and JSON with positions | Coming soon |
MinerU opendatalab/mineru | $2.00 | $200.00 | Markdown and JSON with positions | Coming soon |
Docling ibm/docling | $2.00 | $200.00 | Markdown and JSON with positions | Coming soon |
OCR APIs
Dedicated OCR and document parsing services.
| Model | 1,000 pages | 100,000 pages | Output | Status |
|---|---|---|---|---|
Qianfan OCR baidu/qianfan-ocr | $1.00 | $100.00 | Markdown and JSON | Available |
Google Document AI google/document-ai | $3.00 | $300.00 | Markdown and JSON with positions | Available |
LlamaParse llamaindex/llamaparse | $5.00 | $500.00 | Markdown and JSON with positions | Coming soon |
Mistral OCR mistral/mistral-ocr | $6.00 | $600.00 | Markdown and JSON with positions | Available |
Amazon Textract aws/textract | $6.00 | $600.00 | Markdown and JSON with positions | Available |
Mathpix Convert mathpix/convert | $7.00 | $700.00 | Markdown and JSON with positions | Coming soon |
Marker datalab/marker | $14.00 | $1,400.00 | Markdown and JSON with positions | Coming soon |
Azure Document Intelligence azure/document-intelligence | $15.00 | $1,500.00 | Markdown and JSON with positions | Available |
Reducto Parse reducto/parse | $20.00 | $2,000.00 | Markdown and JSON with positions | Coming soon |
Vision models
General vision models, prompted to transcribe.
| Model | 1,000 pages | 100,000 pages | Output | Status |
|---|---|---|---|---|
Gemini Flash google/gemini-flash | $3.00 | $300.00 | Markdown and JSON | Available |
GPT openai/gpt | $11.00 | $1,100.00 | Markdown and JSON | Available |
Claude anthropic/claude | $14.00 | $1,400.00 | Markdown and JSON | Available |
Prices are what you pay on OCRRouter, checked 2026-11-13. Models marked Coming soon show an estimate. Output describes what the model returns, not how accurate it is.
Let the router pick the cheapest model
You do not have to choose a row from the table. With route set to cheapest, each request goes to the lowest-priced model that is up. If that model fails, the next one answers and you pay only for that one.
Sending the same file to the same model twice is answered from cache at no cost.
Request body
{
"model": "auto",
"route": "cheapest",
"file_url": "https://example.com/document.pdf",
"format": "markdown"
}