GLM-OCR vs OpenOCR
Pick GLM-OCR for high-volume PDF to Markdown, academic papers with formulas, and schema-based field extraction, where a $0.03 per million token API or self-hosting both work. Pick OpenOCR for lightweight self-hosted parsing of Chinese and English documents under Apache-2.0. The main split is scope: GLM-OCR is a hosted and self-hosted document model with a higher OmniDocBench v1.5 score (94.62 vs 90.57), while OpenOCR is a small self-host-only toolkit limited to two languages.
GLM-OCRzai/glm-ocr | OpenOCRfudan/openocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 6 credits / page | 1 credit / page |
| Per 1,000 pages | $6.00 | $1.00 |
| Typical speed | 1.3 s / page | 0.5 s / page |
| Best for | Low-cost general documents | Lightweight OCR pipeline on CPU |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | Chinese and English |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | No | No |
| Multilingual | ✓ Yes | No |
| 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, with self-hosting as an option through vLLM, SGLang, Ollama or transformers.
- You need Markdown or JSON output, including extraction of fields into a user-defined JSON schema from forms.
- You want the higher OmniDocBench v1.5 score: 94.62 against 90.57 for OpenDoc-0.1B, both reported by their authors.
Choose OpenOCR if
- You only handle Chinese and English documents and want a 0.1B model that reports 90.57 on OmniDocBench v1.5.
- You want a plain Apache-2.0 licence and a Python library you run yourself, with ONNX export for deployment.
- You need LaTeX output for formulas, or bounding boxes, from a model that reads plain text and single or multi-line formulas in one pass.
Key differences
| GLM-OCR | OpenOCR | |
|---|---|---|
| Licence | MIT for model weights; SDK repo and bundled PP-DocLayoutV3 are Apache-2.0 | Apache-2.0 |
| How you run it | Z.ai API, or self-host with vLLM, SGLang, Ollama or transformers | Self-host only, as a Python library |
| 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 |
| Model size | 0.9B | 0.1B (OpenDoc-0.1B and UniRec-0.1B) |
| Output formats | Markdown, JSON, plain text | Plain text, LaTeX, bounding boxes |
| OmniDocBench v1.5 overall | 94.62 (authors); also 29.6 mean on ParseBench (independent) | 90.57 for OpenDoc-0.1B (authors); no independent replication found |
| Known failures and limits | Repeats output unless a stop sequence is set; drops table footer rows; loses comma decimal separators; no aarch64 Linux; API caps at 100 pages, 50 MB per PDF, 10 MB per image | Chinese and English only; document parsing needs the separate PP-DocLayoutV2 layout model; no published hardware requirements or throughput |
Sources
- Hugging Face model card
- Z.ai GLM-OCR docs and pricing
- Issue #220 repeated output
- Issue #240 missing table rows
- GitHub README
What each one is
OpenOCR
Fudan's open toolkit for text detection and recognition, formulas, tables and document parsing.
OpenOCR 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": "zai/glm-ocr","fallbacks": ["fudan/openocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
GLM-OCR alternativesOpenOCR alternativesBest OCR models in 2026