PaddleOCR-VL vs GOT-OCR 2.0
Pick PaddleOCR-VL for full-page parsing of multilingual scans, academic papers and report tables, with Markdown and JSON output under an Apache-2.0 licence. Pick GOT-OCR 2.0 for formulas, sheet music, chemical structures and box-guided region OCR on English and Chinese pages. The main split is scope: PaddleOCR-VL parses whole pages in 109 languages, while GOT-OCR 2.0 is a smaller model that outputs LaTeX, TikZ, SMILES and kern.
PaddleOCR-VLpaddle/paddleocr-vl | GOT-OCR 2.0stepfun/got-ocr-2 | |
|---|---|---|
| Type | Open source | Open source |
| Price | 6 credits / page | 1 credit / page |
| Per 1,000 pages | $6.00 | $1.00 |
| Typical speed | 1.3 s / page | 1.5 s / page |
| Best for | CJK documents | Formatted text, formulas and charts |
| Inputs | PDF, images | images |
| Languages | 109 languages | Chinese, English |
| Max pages per request | 500 | 100 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | No | No |
| Multilingual | ✓ Yes | No |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose PaddleOCR-VL if
- It supports 109 languages, including Cyrillic, Arabic, Devanagari and Thai, in one model.
- It outputs Markdown and JSON for full pages, with text, tables, formulas and charts recognized together.
- It has published scores: 96.33 on OmniDocBench v1.6 (PaddleOCR-VL-1.6, by the authors) and 80.0 on olmOCR-Bench (independent).
Choose GOT-OCR 2.0 if
- It outputs LaTeX, TikZ, SMILES and kern, so it covers formulas, chemical structures and sheet music.
- It supports region OCR guided by box coordinates or color, so you can read one part of a page.
- It has 580M parameters, is merged into Hugging Face Transformers with batched inference, and supports dynamic resolution and multi-page OCR.
Key differences
| PaddleOCR-VL | GOT-OCR 2.0 | |
|---|---|---|
| Licence | Apache-2.0 | Code Apache-2.0; training data CC BY-NC 4.0 |
| Size | 0.9B parameters | 580M parameters |
| Output formats | Markdown, JSON | Plain text, Markdown, LaTeX, TikZ, SMILES, kern |
| Languages | 109 languages, including Cyrillic, Arabic, Devanagari and Thai | Best suited to English and Chinese documents |
| Benchmarks | OmniDocBench v1.6: 96.33 (v1.6); olmOCR-Bench: 80.0 | Not published |
| Hardware | NVIDIA GPU recommended (CUDA 12.6 in the install example); CPU possible but slow | CUDA 11.8 and PyTorch 2.0.1 per README; CPU users are pointed to a llama.cpp port |
| Known failures | Version 1.6: Cyrillic regression on born-digital PDFs, use_cache false makes decoding about 11x slower, missing table content, hangs on blurry images | Looping characters mid-recognition, wrong text on blank images instead of empty output, format mode much worse than plain OCR mode |
Sources
- Hugging Face model card
- PaddleOCR-VL-1.6 model card
- olmOCR-Bench table
- Issue #18453 use_cache slowdown
- GOT-OCR2.0 GitHub
- GOT-OCR 2.0 paper
What each one is
PaddleOCR-VL
Open-source document model from Baidu. Excellent on Chinese, Japanese and Korean.
PaddleOCR-VL detailsGOT-OCR 2.0
Unified OCR model that reads plain text, tables, formulas, sheet music and charts.
GOT-OCR 2.0 detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "paddle/paddleocr-vl","fallbacks": ["stepfun/got-ocr-2"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
PaddleOCR-VL alternativesGOT-OCR 2.0 alternativesBest OCR models in 2026