GLM-OCR vs MiniCPM-V 4.5
Pick GLM-OCR for high-volume PDF to Markdown and schema-based field extraction, where its $0.03 per million token API and 0.9B size keep costs low. Pick MiniCPM-V 4.5 for on-device or single-GPU work where you want a general vision model you fully self-host under Apache-2.0. Do not use GLM-OCR for financial tables where every row and decimal must survive.
GLM-OCRzai/glm-ocr | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 6 credits / page | 2 credits / page |
| Per 1,000 pages | $6.00 | $2.00 |
| Typical speed | 1.3 s / page | 2.5 s / page |
| Best for | Low-cost general documents | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | 100+ languages | 30+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | 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 GLM-OCR if
- You want a hosted API at $0.03 per million tokens for input and output, or a free self-hosted 0.9B model.
- You need to extract fields from forms into a user-defined JSON schema, not just transcribe text.
- You process academic papers with formulas or large PDF batches, and the authors report 94.62 on OmniDocBench v1.5 and 1.86 pages per second on one replica.
Choose MiniCPM-V 4.5 if
- You need to run OCR on-device or at the edge, using int4, GGUF or AWQ builds on llama.cpp or vLLM.
- You want local document Q&A on a single GPU from one general vision language model.
- You want Apache-2.0 on both weights and code, with no hosted service involved.
Key differences
| GLM-OCR | MiniCPM-V 4.5 | |
|---|---|---|
| Model size | 0.9B, layout analysis plus a GLM decoder | 8.7B, general vision language model on Qwen3-8B and SigLIP2-400M |
| How you run it | Z.ai API, or self-host with vLLM, SGLang, Ollama or transformers | Self-host only, with llama.cpp, vLLM, Ollama or Transformers (Ollama and SGLang were still pending official releases on the model card) |
| Licence | MIT for weights; SDK repo and bundled PP-DocLayoutV3 are Apache-2.0 | Apache-2.0 for weights and code |
| Price | API: $0.03 per million tokens for input and output. Self-hosted: free, you pay for compute | Free, self-hosted, you pay for compute |
| Benchmark evidence | OmniDocBench v1.5 overall 94.62 (authors), olmOCR-Bench 75.2 (authors), ParseBench mean 29.6 (independent) | Authors claim leading OCRBench and state of the art PDF parsing among general MLLMs on OmniDocBench, but the model card gives scores only as images |
| Output formats | Markdown, JSON, plain text; can fill a user-defined JSON schema | Plain text, Markdown, JSON; no bounding box or layout output |
| Known failure modes and limits | Repeats output unless a stop sequence is set, drops table footer rows, loses comma decimals in cropped tables, does not run on aarch64 Linux; API caps at 100 pages, 50 MB per PDF, 10 MB per image | No int4 VRAM figure published; no hosted SLA |
Sources
- GLM-OCR Hugging Face model card
- Z.ai GLM-OCR docs and pricing
- GLM-OCR issue #240 missing table rows
- MiniCPM-V 4.5 Hugging Face model card
- MiniCPM-V GitHub repo
What each one is
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 detailsTry 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": "zai/glm-ocr","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
GLM-OCR vs DeepSeek-VL2GLM-OCR vs InternVL3.5Unlimited OCR vs MiniCPM-V 4.5PaddleOCR-VL vs MiniCPM-V 4.5Jina OCR vs GLM-OCRTrOCR vs MiniCPM-V 4.5GLM-OCR vs Tesseract.jsGLM-OCR vs Nanonets-OCR
GLM-OCR alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026