Google Document AI vs MiniCPM-V 4.5
Pick MiniCPM-V 4.5 if you want to run OCR yourself under Apache-2.0, for example on a single GPU, on-device, or for local document Q&A. Pick Google Document AI if you run on Google Cloud and need a hosted API that returns bounding boxes, chunks and batch processing of up to 500 pages. The main split is deployment: Document AI is a paid, proprietary service, while MiniCPM-V is free weights you must host, and it gives no coordinates.
Google Document AIgoogle/document-ai | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| Type | OCR API | Vision model |
| Price | 10 credits / page | 2 credits / page |
| Per 1,000 pages | $10.00 | $2.00 |
| Typical speed | 1.4 s / page | 2.5 s / page |
| Best for | Multilingual scans | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | 200+ languages | 30+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | No | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | No |
| Forms | ✓ Yes | 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 Google Document AI if
- You need bounding boxes in the output, which MiniCPM-V does not provide as a general VLM without layout output.
- You want RAG chunking from Layout Parser at $10 per 1,000 pages, instead of building your own splitter.
- You process bulk volume on Google Cloud: Enterprise OCR is $1.50 per 1,000 pages, drops to $0.60 above 5M pages a month, and batch requests take up to 500 pages.
Choose MiniCPM-V 4.5 if
- You want open weights and code under Apache-2.0, with no per-page fee beyond your own compute.
- You need on-device or edge OCR, since int4, GGUF and AWQ builds run on llama.cpp and vLLM.
- You want local document Q&A on a single GPU, with output as plain text, Markdown or JSON.
Key differences
| Google Document AI | MiniCPM-V 4.5 | |
|---|---|---|
| How you run it | Hosted API on Google Cloud | Self-host with llama.cpp, vLLM, Ollama or Transformers |
| Licence | Proprietary | Apache-2.0 weights and code |
| Price | Enterprise OCR $1.50 per 1,000 pages ($0.60 above 5M a month, first 1,000 free); Layout Parser $10; Form Parser $30 per 1,000 pages | Free, you pay for compute |
| Output formats | JSON, plain text, bounding boxes, chunks | Plain text, Markdown, JSON; no bounding box or layout output |
| Request limits | Online requests up to 15 pages (30 in imageless mode); batch up to 500 pages; Form Parser batch up to 100 pages | Not published |
| Benchmark | 64.6% on RD-TableBench table accuracy, behind Azure and Textract | Not published as text; OpenBMB claims leading OCRBench and OmniDocBench PDF parsing results, with scores given only as images |
| Known weak spot | Complex tables and long PDFs in one synchronous call | Ollama and SGLang support were still pending on the model card; no int4 VRAM figure published |
Sources
What each one is
Google Document AI
Google Cloud OCR with layout parsing and specialised processors.
Google Document AI 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": "google/document-ai","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Google Document AI alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026