PaddleOCR-VL 1.5
PaddleOCR-VL 1.5 is Baidu PaddlePaddle's 0.9B document model, an update to PaddleOCR-VL with better tables, formulas and photographed pages.
A release of PaddleOCR-VL.
- Released
- January 29, 2026
- Model ID
- PaddlePaddle/PaddleOCR-VL-1.5, PaddlePaddle/PaddleOCR-VL-1.5-GGUF
- Size
- 0.9B
- License
- Apache-2.0
- Also searched as
- paddleocr-vl-1.5, paddleocr vl 1.5
What is new in PaddleOCR-VL 1.5
- 94.5% on OmniDocBench v1.5, above PaddleOCR-VL, with gains on tables, formulas and text.
- Finds text in irregular shapes, for skewed, warped, scanned or photographed pages and uneven light.
- Adds text spotting and seal recognition.
- Better rare characters, ancient texts, multilingual tables, underlines and checkboxes. Adds Tibetan and Bengali.
- Merges tables and headings that continue across pages.
PaddleOCR-VL 1.5 benchmarks
| Benchmark | Score | Note |
|---|---|---|
| OmniDocBench v1.5, overall | 94.5% | Above PaddleOCR-VL |
Scores as published by the authors. We have not rerun them.
Use PaddleOCR-VL 1.5 through an API
OCRRouter runs PaddleOCR-VL 1.6, the release after 1.5, as paddle/paddleocr-vl.
curl https://api.ocrrouter.ai/v1/ocr \
-H "Authorization: Bearer $OCRROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "paddle/paddleocr-vl",
"file_url": "https://example.com/invoice.pdf",
"format": "markdown"
}'PaddleOCR-VL versions
- PaddleOCR-VL: all versions, price, speed and features on OCRRouter
- PaddleOCR-VL 1.5 (January 29, 2026), this page
Questions
Sources
Last updated .