OCRRouter

GLM-OCR vs RapidOCR

Pick GLM-OCR when you need Markdown, tables, formulas or JSON field extraction from PDFs, and you can accept its table and repeat-output bugs. Pick RapidOCR when you only need plain text and boxes from Chinese or English images, offline on CPU, in Python, C++, Java or C#. The main split is output: GLM-OCR parses documents into structured text, RapidOCR returns text lines and coordinates only.

GLM-OCRzai/glm-ocrRapidOCRrapidai/rapidocr
TypeOpen sourceOpen source
Price6 credits / page1 credit / page
Per 1,000 pages$6.00$1.00
Typical speed1.3 s / page0.6 s / page
Best forLow-cost general documentsLightweight OCR on CPU or edge
InputsPDF, imagesimages
Languages100+ languagesChinese, English and more
Max pages per request500100
Tables✓ YesNo
HandwritingNoNo
Math✓ YesNo
Bounding boxes✓ Yes✓ Yes
Confidence✓ Yes✓ Yes
FormsNoNo
Multilingual✓ Yes✓ Yes
On OCRRouterAvailableAvailable

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 outputs Markdown, JSON and plain text, and can extract fields into a user-defined JSON schema.
  • The Z.ai API costs $0.03 per million tokens for input and output, and the weights are MIT-licensed for self-hosting.
  • It scores 94.62 on OmniDocBench v1.5 in its model card, and suits academic papers with formulas.

Choose RapidOCR if

  • It runs offline on CPU through ONNX Runtime, with optional TensorRT, OpenVINO or MNN engines.
  • It has bindings for Python, C++, Java and C#, so you can embed it in Java or C# apps.
  • It is free under Apache-2.0 and fits Chinese and English screenshots and receipts.

Key differences

GLM-OCRRapidOCR
OutputMarkdown, JSON, plain textPlain text and bounding boxes
How you run itZ.ai API, or self-host with vLLM, SGLang, Ollama or transformersLocal library for Python, C++, Java or C#
LicenceMIT for weights; SDK repo and PP-DocLayoutV3 are Apache-2.0Apache-2.0; converted models keep their upstream licences
PriceAPI: $0.03 per million tokens. Self-hosted: free, you pay for computeFree, self-hosted, you pay for compute
BenchmarkOmniDocBench v1.5: 94.62 (authors). ParseBench mean: 29.6 (independent)Not published
LanguagesNot publishedChinese and English by default; other languages need extra models
Known failuresRepeated output without a stop sequence, dropped table footer rows, lost comma decimals, no aarch64 Linux support, API limit of 100 pages, 50 MB per PDF, 10 MB per image16-bit grayscale PNG or TIFF give no or wrong text, sliced inference crashes on large pages, Arabic recognition crashes

Sources

What each one is

GLM-OCR

Small open-source OCR model from Z.ai with layout regions. MIT licensed.

GLM-OCR details

RapidOCR

PaddleOCR models packaged for ONNX Runtime and other engines. Easy to deploy without Paddle.

RapidOCR details

Try 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": ["rapidai/rapidocr"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

GLM-OCR alternativesRapidOCR alternativesBest OCR models in 2026