dots.ocr vs GOT-OCR 2.0
Pick dots.ocr for multilingual PDFs that need layout boxes, HTML tables and Markdown in one pass under an MIT licence. Pick GOT-OCR 2.0 for formulas, sheet music, chemical structures or OCR of a chosen region, and for a smaller 580M model. The main split is scope: dots.ocr parses whole pages, GOT-OCR 2.0 reads many content types but publishes no benchmark scores.
dots.ocrrednote/dots-ocr | GOT-OCR 2.0stepfun/got-ocr-2 | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 1.5 s / page | 1.5 s / page |
| Best for | Multilingual document parsing | Formatted text, formulas and charts |
| Inputs | PDF, images | images |
| Languages | 100+ languages | Chinese, English |
| Max pages per request | 500 | 100 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | 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 dots.ocr if
- It returns layout boxes, categories and text in one pass, with tables as HTML and formulas as LaTeX.
- The authors publish scores: 79.1 on olmOCR-bench for the original and 83.9 for dots.mocr (v1.5).
- The code is MIT-licensed, while GOT-OCR 2.0's training data is CC BY-NC 4.0.
Choose GOT-OCR 2.0 if
- It reads formulas, sheet music, chemical structures (SMILES) and geometry, with outputs such as LaTeX, TikZ and kern.
- Region OCR works from box coordinates or color, so you can read one part of a page.
- At 580M parameters it is much smaller than the 3B dots.ocr, and it is merged into Hugging Face Transformers with batched inference.
Key differences
| dots.ocr | GOT-OCR 2.0 | |
|---|---|---|
| Licence | MIT (repository also includes a separate dots.ocr license agreement file) | Code Apache-2.0; training data CC BY-NC 4.0 |
| Size | 3B (1.7B LLM backbone) | 580M |
| How you run it | Self-host. Single CUDA GPU with vLLM is the documented path; Hugging Face Transformers also works. No VRAM figure stated. | Self-host in Hugging Face Transformers. README lists CUDA 11.8 and PyTorch 2.0.1; CPU users are pointed to a llama.cpp port. |
| Price | Free, you pay for compute | Free, you pay for compute |
| Output formats | JSON, Markdown, HTML, LaTeX, bounding boxes, SVG | Plain text, Markdown, LaTeX, TikZ, SMILES, kern |
| Published benchmarks | olmOCR-bench 79.1 (dots.ocr), 83.9 (dots.mocr); OmniDocBench v1.5 text edit distance 0.031 (dots.mocr, lower is better). Scores are from the authors. | Not published |
| Known failure | Runs of ellipses or underscores can make output repeat endlessly. A table above 1000 grid cells raises an error and drops the whole page (Docling issue). | Output can loop characters midway, and images with no text can return wrong text instead of an empty string. Format mode gives much worse results than plain OCR mode. |
Sources
What each one is
dots.ocr
Single vision model from rednote that does layout detection and content recognition together.
dots.ocr 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": "rednote/dots-ocr","fallbacks": ["stepfun/got-ocr-2"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
dots.ocr alternativesGOT-OCR 2.0 alternativesBest OCR models in 2026