Jina OCR vs MiniCPM-V 4.5
Pick Jina OCR for non-commercial PDF to Markdown work, where it scores 91.14 on OmniDocBench v1.6 (reported by its authors) and activates only about 570M parameters. Pick MiniCPM-V 4.5 if you ship a commercial product or need quantized, on-device OCR, because its Apache-2.0 licence allows commercial use while Jina's CC-BY-NC-4.0 does not without a separate license. MiniCPM-V 4.5 also publishes no verifiable benchmark numbers, so test it on your own documents.
Jina OCRjina/jina-ocr | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 1.5 s / page | 2.5 s / page |
| Best for | Clean Markdown at low cost | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | About 100 languages | 30+ languages |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | No | 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 Jina OCR if
- It scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, 7.4 points above DeepSeek-OCR. These are author-reported numbers.
- It has only about 570M active parameters (3B MoE) and runs at 2.57 pages per second on an A100.
- It can output Markdown, HTML and LaTeX, and you can either self-host it or call Jina's hosted API.
Choose MiniCPM-V 4.5 if
- Apache-2.0 weights and code allow commercial use with no extra license.
- It runs on llama.cpp and vLLM with int4, GGUF and AWQ builds, which suits on-device or edge OCR and local document Q&A on a single GPU.
- As a general vision language model it outputs plain text, Markdown and JSON, so you can also ask questions about a document, not just convert it.
Key differences
| Jina OCR | MiniCPM-V 4.5 | |
|---|---|---|
| Licence | CC-BY-NC-4.0; commercial use needs a license from Jina/Elastic | Apache-2.0 |
| How you run it | Self-host or Jina's hosted API | Self-host only (llama.cpp, vLLM, Transformers; Ollama and SGLang support was still pending) |
| Model size | 3B MoE, about 570M active | 8.7B dense VLM (Qwen3-8B plus SigLIP2-400M) |
| Output formats | Markdown, HTML, LaTeX | Plain text, Markdown, JSON |
| Published benchmarks | OmniDocBench v1.6 91.14; olmOCR-Bench 83.4 (authors) | Not published as text; scores appear only as images on the model card |
| Hardware | GPU recommended (A100, L4 benchmarked); vLLM 0.21+ for FastMTP; CPU fallback in Transformers | Quantized builds available; no int4 VRAM figure published |
| Known failure | Hosted API returns HTTP 503 on cold start; retry after 30 to 60 seconds. Reported scores use a stricter prompt than the default. | No bounding box or layout output, so it is not suited to jobs that need coordinates |
Sources
What each one is
Jina OCR
Compact page-to-Markdown model from Jina AI, built on the DeepSeek-OCR encoder.
Jina OCR 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": "jina/jina-ocr","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Jina OCR alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026