dots.ocr vs Phi-4-multimodal
Pick dots.ocr for multilingual PDFs where you need layout, bounding boxes and tables as HTML in one pass. Pick Phi-4-multimodal for English document Q&A, or when your pipeline also handles audio. The main split is language and output: Phi-4-multimodal supports vision in English only and returns no coordinates, while dots.ocr returns layout boxes.
dots.ocrrednote/dots-ocr | Phi-4-multimodalmicrosoft/phi-4-multimodal | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 1.5 s / page | 2.5 s / page |
| Best for | Multilingual document parsing | Small model for OCR and questions |
| Inputs | PDF, images | images |
| Languages | 100+ languages | English for vision |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | No |
| 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 is built for multilingual documents, while Phi-4-multimodal supports vision for English only.
- It returns layout boxes, categories and text in one pass, with tables as HTML and formulas as LaTeX.
- It is a 3B model, and the dots.mocr (v1.5) release scores 83.9 on olmOCR-bench, up from 79.1 for the original.
Choose Phi-4-multimodal if
- It scores 93.2 on DocVQA and 84.4 on OCRBench per Microsoft, which suits English document question answering.
- It takes audio as well as images, so one model covers mixed audio and image pipelines.
- It has a 128K token context and can be used through Azure AI Foundry as well as self-hosted.
Key differences
| dots.ocr | Phi-4-multimodal | |
|---|---|---|
| Licence | MIT (the repository also includes a separate dots.ocr license agreement file) | MIT |
| How you run it | Self-host only, with vLLM or Hugging Face Transformers | Self-host or use it via Azure AI Foundry |
| Size | 3B (1.7B LLM backbone) | 5.6B |
| Output formats | JSON, Markdown, HTML, LaTeX, bounding boxes, SVG | Plain text, Markdown, JSON |
| Languages | Multilingual | Vision is English only, and performance drops in other languages |
| Hardware | Single CUDA GPU with vLLM is the documented path. No VRAM figure is stated. | Tested on A100, A6000 and H100. V100 and older need eager attention. |
| Known failure | Runs of ellipses or underscores can make output repeat endlessly. Very large tables are hard, and a table above 1000 grid cells drops the page in Docling. | The model card warns of repetitive or inconsistent output in long chats. |
Sources
What each one is
dots.ocr
Single vision model from rednote that does layout detection and content recognition together.
dots.ocr detailsPhi-4-multimodal
Microsoft's 5.6B text, vision and audio model (February 2025). Scores 84.4 on OCRBench per its model card.
Phi-4-multimodal 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": ["microsoft/phi-4-multimodal"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
dots.ocr alternativesPhi-4-multimodal alternativesBest OCR models in 2026