GLM-OCR vs PaddleOCR (PP-OCR)
Pick GLM-OCR when you need whole documents converted to Markdown or JSON, such as academic papers with formulas or forms with field extraction, and you want a cheap API at $0.03 per million tokens. Pick PaddleOCR (PP-OCR) when you need fast, small, local text detection and recognition for printed Chinese and English, receipts, signs or screenshots, including on CPU, edge devices or in the browser. The main split is output: GLM-OCR returns structured documents, PP-OCR returns text lines and boxes.
GLM-OCRzai/glm-ocr | PaddleOCR (PP-OCR)paddle/paddleocr | |
|---|---|---|
| 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 | 0.7 s / page |
| Best for | Low-cost general documents | Fast printed text, Chinese and English |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | 80+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | ✓ Yes |
| 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
- It converts PDFs to Markdown or JSON, and can extract fields into a user-defined JSON schema.
- It is available as a Z.ai API at $0.03 per million tokens for input and output, or self-hosted with vLLM, SGLang, Ollama or transformers.
- It scores 94.62 on OmniDocBench v1.5 per its model card, and the authors report 1.86 PDF pages per second on a single replica.
Choose PaddleOCR (PP-OCR) if
- It is a 34.5M parameter model (PP-OCRv6 medium) that runs on CPU, NVIDIA GPU, Kunlunxin XPU and other accelerators, with a 5.2x CPU speedup using OpenVINO.
- One PP-OCRv6 model covers 50 languages, including Chinese, English, Japanese and 46 Latin-script languages.
- It runs in the browser through the official PaddleOCR.js SDK, and in C++ or Python, under Apache-2.0.
Key differences
| GLM-OCR | PaddleOCR (PP-OCR) | |
|---|---|---|
| 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 | Local only: paddleocr Python package, C++ or PaddleOCR.js in the browser |
| 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 parameters | 34.5M parameters (PP-OCRv6 medium) |
| Output | Markdown, JSON, plain text | Plain text, bounding boxes, JSON. Tables and formulas need the separate PP-Structure or VL pipelines |
| Benchmarks | OmniDocBench v1.5: 94.62 (authors). ParseBench mean: 29.6 (independent) | Not published as a benchmark score. Authors report PP-OCRv6 medium beats PP-OCRv5_server by 4.6% detection and 5.1% recognition |
| Known failures | Repeated output unless a stop sequence is set; drops table footer and summary rows; loses comma decimal separators in cropped tables; no aarch64 Linux support; API caps at 100 pages, 50 MB per PDF, 10 MB per image | PP-OCRv6 weights contain denormals that make CPU inference 4 to 11x slower; NotImplementedError on PaddleOCR 3.7.0 with PaddlePaddle 3.3.1; training skips the last batch of every epoch on Windows |
Sources
- Hugging Face model card
- Z.ai GLM-OCR docs and pricing
- Issue #240 missing table rows
- PaddleOCR GitHub README
- PP-OCRv6 on Hugging Face
- RapidOCR issue #743 denormal weights
What each one is
PaddleOCR (PP-OCR)
Baidu's classic open-source text detection and recognition pipeline. Light enough to run on CPU.
PaddleOCR (PP-OCR) 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": ["paddle/paddleocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
GLM-OCR alternativesPaddleOCR (PP-OCR) alternativesBest OCR models in 2026