Unlimited OCR vs SmolDocling
Pick Unlimited OCR for long reports and books that you want parsed in one request on an NVIDIA GPU, with Markdown, plain text and bounding boxes out. Pick SmolDocling only for experiments or cheap English page conversion on small hardware, because it is a 256M preview release and its authors point users to Granite-Docling.
Unlimited OCRbaidu/unlimited-ocr | SmolDoclinghuggingface/smoldocling | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 1.2 s / page | 1.0 s / page |
| Best for | Books and very long documents | Very cheap page conversion |
| Inputs | PDF, images | PDF, images |
| Languages | 30+ languages | Mainly English |
| Max pages per request | 1,000 | 500 |
| 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 Unlimited OCR if
- It parses 40+ pages in a single forward pass with roughly constant memory, so a whole report or book goes in one request.
- It is MIT licensed, which allows commercial use with no copyleft, and Baidu Cloud also offers a hosted endpoint.
- Unlimited OCR 2 scores 77.1 on MarathonDocBench versus 33.7 for the next model, per Baidu.
Choose SmolDocling if
- At 256M parameters it is small enough for modest hardware, and it runs at about 0.35 seconds per page on an A100 with vLLM.
- One model handles text, tables (OTSL), code, formulas and charts with location tags, and it can output DocTags, Markdown, HTML, JSON and LaTeX.
- It is a low-cost way to experiment with a tiny document VLM on English pages.
Key differences
| Unlimited OCR | SmolDocling | |
|---|---|---|
| Licence | MIT | Apache-2.0 per the model summary (the metadata tag says CDLA-Permissive-2.0) |
| How you run it | Self-host with vLLM, SGLang or transformers, or use the hosted endpoint in Baidu Cloud's OCR product | Self-host only |
| Model size | 3B total, about 500M active per token (Unlimited OCR 2 is 3.34B) | 256M |
| Hardware | NVIDIA GPU with CUDA 12.9 or 13.0; version 2 under SGLang runs out of memory on 12 GB GPUs; not for CPU-only machines | Small enough for modest hardware |
| Output formats | Markdown, plain text, bounding boxes | DocTags, Markdown, HTML, JSON, LaTeX, bounding boxes |
| Languages | Not published | English only |
| Known failure | Text repetition and looping (issue #55), hallucinated text (issue #58), bbox coordinate loops with SGLang on GB10 GPUs (issue #114), and styling such as bold and italic is mostly lost | Endless token repetition and missing closing tags; output loops on some images, making generation very slow (Docling issue #1515) |
Sources
- Hugging Face model card
- MarkTechPost on Unlimited OCR 2
- Issue #116 OOM on 12 GB
- SmolDocling model card
- SmolDocling paper
- Docling issue #1515
What each one is
Unlimited OCR
Open-source model from Baidu that reads hundreds of pages in one pass with flat memory use.
Unlimited OCR detailsSmolDocling
Tiny 256M vision model from Hugging Face and IBM that outputs DocTags. Predecessor of Granite Docling.
SmolDocling detailsTry both on your own file
Run the file once with each model and compare the output.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "baidu/unlimited-ocr","fallbacks": ["huggingface/smoldocling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Unlimited OCR vs MonkeyOCRUnlimited OCR vs Granite DoclingMinerU vs SmolDoclingGLM-OCR vs SmolDoclingDeepSeek-OCR vs Unlimited OCREasyOCR vs SmolDoclingMistral OCR vs SmolDoclingAzure Document Intelligence vs Unlimited OCR
Unlimited OCR alternativesSmolDocling alternativesBest OCR models in 2026