Surya vs MiniCPM-V 4.5
Pick Surya when you need layout labels, reading order and bounding boxes, or want OCR on a laptop with 4-6 GB of RAM. Pick MiniCPM-V 4.5 when you want a general vision language model under a plain Apache-2.0 licence for local document Q&A or edge use. The main split is output: Surya returns coordinates and structure, MiniCPM-V returns text, Markdown or JSON only.
Suryadatalab/surya | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| 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 | Layout and line-level OCR | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | 90+ languages | 30+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | No | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| 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 Surya if
- It returns layout labels, reading order, HTML tables, KaTeX math and per-block bounding boxes from one model.
- It runs on CPU and Apple Silicon through llama.cpp, with peak memory of about 4.3-5.9 GB on an M4 Max.
- Its authors report 83.7% on olmOCR-bench with only 689M parameters, and 4.52 pages/s at 192 DPI on one H100 with vLLM.
Choose MiniCPM-V 4.5 if
- It is Apache-2.0 for both code and weights, with no revenue or funding cap.
- It is a general VLM, so you can run local document Q&A on a single GPU, not just extraction.
- It ships int4, GGUF and AWQ builds and runs on llama.cpp and vLLM, which suits on-device and edge jobs.
Key differences
| Surya | MiniCPM-V 4.5 | |
|---|---|---|
| Licence | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M, free for research, personal use and companies under $5M funding or revenue | Apache-2.0 for weights and code |
| Cost | Free self-hosted; commercial licence needed above $5M revenue or funding | Free self-hosted; you pay for compute |
| Model size | 689M parameters | 8.7B parameters, built on Qwen3-8B and SigLIP2-400M |
| Output formats | JSON, HTML, LaTeX, bounding boxes, plain text | Plain text, Markdown, JSON; no bounding box or layout output |
| How you run it | Light client plus a separate vLLM or llama.cpp inference server | llama.cpp, vLLM, Ollama or Transformers; Ollama and SGLang support were still pending on the model card |
| Benchmarks | 83.7% on olmOCR-bench and 85.9 on OmniParseBench, reported by the authors | OpenBMB claims leading OCRBench results, but the model card shows scores only as images |
| Known weak spots | Handwriting, complex multilingual tables (58.1 score) and photos or natural scenes; Surya 2.5 tables lack row and column geometry | No int4 VRAM figure published; not suited to jobs that need coordinates |
Sources
What each one is
Surya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya detailsMiniCPM-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 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": "datalab/surya","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Surya alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026