OCRRouter

dots.ocr vs MiniCPM-V 4.5

Pick dots.ocr when you need layout-aware parsing of multilingual PDFs, with bounding boxes, tables as HTML and formulas as LaTeX from a single 3B model. Pick MiniCPM-V 4.5 when you want a general 8.7B vision language model for local document Q&A or quantized edge OCR, and do not need coordinates. The main split is structured layout output versus a general VLM.

dots.ocrrednote/dots-ocrMiniCPM-V 4.5openbmb/minicpm-v
TypeOpen sourceVision model
Price2 credits / page2 credits / page
Per 1,000 pages$2.00$2.00
Typical speed1.5 s / page2.5 s / page
Best forMultilingual document parsingOCR on edge devices
InputsPDF, imagesimages
Languages100+ languages30+ languages
Max pages per request500500
Tables✓ Yes✓ Yes
HandwritingNo✓ Yes
Math✓ YesNo
Bounding boxes✓ YesNo
ConfidenceNoNo
FormsNoNo
Multilingual✓ Yes✓ Yes
On OCRRouterAvailableAvailable

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

  • You need bounding boxes, layout categories and text in one pass, which MiniCPM-V 4.5 does not output.
  • You convert mixed text and table documents to Markdown, HTML or LaTeX, and want published olmOCR-bench scores (79.1 for dots.ocr, 83.9 for dots.mocr).
  • You want a smaller 3B model, or the dots.mocr-svg variant to turn charts and diagrams into SVG.

Choose MiniCPM-V 4.5 if

  • You want Apache-2.0 weights and code with no separate licence agreement file to review.
  • You need int4, GGUF or AWQ builds that run on llama.cpp or vLLM for on-device or edge use.
  • You want one model for local document Q&A on a single GPU, not only parsing.

Key differences

dots.ocrMiniCPM-V 4.5
Model size3B (1.7B LLM backbone)8.7B, built on Qwen3-8B and SigLIP2-400M
LicenceMIT, plus a separate dots.ocr license agreement file in the repositoryApache-2.0 for weights and code
How you run itSelf-hosted with vLLM or Hugging Face Transformers. Single CUDA GPU with vLLM is the documented path. No VRAM figure stated.Self-hosted with llama.cpp, vLLM, Ollama or Transformers. Int4, GGUF and AWQ builds exist. No int4 VRAM figure published.
Output formatsJSON, Markdown, HTML, LaTeX, bounding boxes, SVGPlain text, Markdown, JSON. No bounding box or layout output.
BenchmarksolmOCR-bench 79.1 (dots.ocr) and 83.9 (dots.mocr), reported by the authors. OmniDocBench v1.5 text edit distance 0.031 for dots.mocr.Not published as text. The model card shows scores only as images, so they are hard to verify. OpenBMB claims leading OCRBench results.
Known failureRuns of ellipses or underscores can make output repeat endlessly. A table above 1000 grid cells raised an error and dropped the whole page in Docling.Ollama and SGLang support were still pending official releases on the model card.
PriceFree, self-hosted. You pay for compute.Free, self-hosted. You pay for compute.

Sources

What each one is

dots.ocr

Single vision model from rednote that does layout detection and content recognition together.

dots.ocr details

MiniCPM-V 4.5

8B vision model from OpenBMB (August 2025) built to run on devices, with strong OCR and document parsing.

MiniCPM-V 4.5 details

Try 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": "rednote/dots-ocr","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

dots.ocr alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026