Unlimited OCR vs Florence-2
Pick Unlimited OCR to turn long reports and books into text with boxes in one request, for example bulk PDF to text for RAG. Pick Florence-2 for short English text in images, such as scene text and labels, on small hardware or in a pipeline that also needs detection. The main split is scope: Unlimited OCR reads 40+ pages in one pass, while Florence-2 returns only text and boxes and has no document structure.
Unlimited OCRbaidu/unlimited-ocr | Florence-2microsoft/florence-2 | |
|---|---|---|
| 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 | 0.8 s / page |
| Best for | Books and very long documents | Lightweight OCR with boxes |
| Inputs | PDF, images | images |
| Languages | 30+ languages | English |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| 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 long report or book goes in as one request.
- It emits block types and bounding boxes alongside text, and outputs Markdown, which suits layout-aware extraction and RAG ingestion.
- Baidu reports 77.1 for Unlimited OCR 2 on MarathonDocBench versus 33.7 for the next model, and Baidu Cloud also offers a hosted endpoint if you do not want to run GPUs.
Choose Florence-2 if
- It is very small, at 0.23B (base) and 0.77B (large), which fits lightweight edge OCR in English.
- OCR_WITH_REGION returns a quadrilateral box for each text line, which works well for scene text and short labels.
- The same model also does captioning, detection and grounding, so one model can cover a pipeline that needs more than OCR.
Key differences
| Unlimited OCR | Florence-2 | |
|---|---|---|
| Licence | MIT | MIT |
| How you run it | Self-host with vLLM, SGLang or transformers, or use the hosted endpoint in Baidu Cloud's OCR product | Self-host via Transformers only |
| Model size | 3B total (about 500M active per token); Unlimited OCR 2 is 3.34B | 0.23B (base), 0.77B (large) |
| Output formats | Markdown, plain text, bounding boxes | Plain text and bounding boxes. No Markdown, tables or LaTeX |
| 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 | Not published |
| Page scope | 40+ pages in one pass, but context is capped at 32,768 tokens and multi-page input only works in base mode | Not suited to full-page document conversion |
| Known failures | Text repetition and looping on some pages, reports of hallucinated text, bbox coordinate loops with SGLang on GB10 GPUs, and text styling mostly lost | Spelling errors and missing spaces on cursive handwriting and French text, mostly wrong Chinese output because the vocabulary lacks CJK characters |
Sources
- Hugging Face model card (Unlimited OCR)
- GitHub repo (Unlimited OCR)
- MarkTechPost on Unlimited OCR 2
- Florence-2-large model card
- HF blog OCR test
- HF discussion Chinese very bad
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 detailsFlorence-2
Small Microsoft vision model (0.23B and 0.77B) with OCR and OCR-with-regions tasks that return text with boxes.
Florence-2 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": "baidu/unlimited-ocr","fallbacks": ["microsoft/florence-2"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Unlimited OCR alternativesFlorence-2 alternativesBest OCR models in 2026