OCR
PaddleOCR-VL 1.6
Compact multilingual document OCR for text, tables, formulas, charts, and seals.
- Status
- Live
- Languages
- Multilingual
- License
- Apache 2.0 (PaddlePaddle)
Model ID
| Model ID | Designed for |
|---|---|
paddleocr-vl-1.6 | Page-level document parsing with Markdown and structured blocks |
Confirm access with GET /v1/models, then follow the utility-model docs for the canonical v1 REST or native gRPC request contract.
Pricing
paddleocr-vl-1.6: Free for a limited time. $0.001 $0.00 per successful page
Overview
PaddleOCR-VL 1.6 replaces Unlimited OCR as Tarka’s page-level document parser.
The paddleocr-vl-1.6 alias runs PaddlePaddle’s official v1.6
pipeline, including PP-DocLayoutV3, through POST /v1/ocr.
The portable response remains an ocr.completion with top-level
text. Set output_format to markdown and
request blocks or layout in include when
your application also needs per-page content, reading order, labels, and
geometry.
When to pick it
- A page contains more than ordinary running text.
- Markdown structure matters downstream.
- You need one request to recognize text, tables, formulas, charts, or seals.
- You want page and block data without giving up a simple top-level text value.
When to pick something else
- Nepali or Devanagari-specific extraction → GLM-OCR
- Geometry and reading order without recognition → PP-DocLayoutV3
Send inline base64 image data in image or an ordered
images array. Tarka never fetches a remote image URL on your
behalf. Use GET /v1/models to confirm the alias is enabled for your
API key.