What is GLM-OCR?
Parse PDFs, tables, formulas and handwriting into JSON.
Who is GLM-OCR best for?
GLM-OCR is best for developers who turn scans and PDFs into clean Markdown or JSON.
Document pipelines for RAG
It parses large batches of PDFs and scans into consistent Markdown, which gives a retrieval system clean text to index.
Forms, receipts and ID cards
Key fields from cards, certificates, receipts and forms come out as structured JSON, ready for banking, insurance or logistics workflows.
Tables, formulas and handwriting
Tables come back as HTML, and it reads printed text, handwriting and maths formulas from photos, screenshots and scans.
OCR that stays on your own machines
At 0.9B parameters the open model runs locally through Ollama or MLX, so sensitive documents never have to leave your hardware.
How much does GLM-OCR cost?
GLM-OCR is free to download and run on your own hardware under the MIT License, and Z.ai's hosted API charges $0.03 per million tokens for both input and output.
- Download from Hugging Face or ModelScope
- Runs on vLLM, SGLang, Ollama or MLX on Apple Silicon
- 0.9B parameters, small enough for edge devices
- Hosted parsing, no GPU needed
- PDFs up to 50 MB and 100 pages, images up to 10 MB
- Returns text, Markdown and image links
Input and output tokens cost the same on the API, and there is no cached-input price for GLM-OCR. Z.ai's pricing page lists no free API allowance for this model, so the free route is running the open weights yourself. The SDK's code is under the Apache License 2.0, and its layout step uses PP-DocLayoutV3, also under Apache 2.0. Full pricing on docs.z.ai. We checked these prices on .
How do readers rate GLM-OCR?
No one has rated GLM-OCR yet. Be the first.
Sign in to rate. One rating per account.
Is GLM-OCR open source?
Yes. GLM-OCR's code is public under the Apache 2.0 licence, so you can run it on your own server.
On GitHub
7,594stars
- Last commit
- 14 Oct
- Repository age
- 9 months
- Latest release
- 8 Apr
- Licence
- Apache 2.0
- Self-hosted
- Yes
- Repository
- zai-org/GLM-OCR
What are the best GLM-OCR alternatives?
GLM-OCR is an open source alternative to Cohere Parse 5, Mistral OCR 4.1 and Monkt. Interfaze 1.1 Lite does the same job and is open source too.
Cohere Parse 5FreemiumCohere's document extraction to Markdown and blocks, priced per 1,000 pages.
Mistral OCR 4.1FreeMistral's OCR with bounding boxes, block labels and confidence scores.
MonktPaidPDF, Word and Excel into AI-ready Markdown or JSON.
Interfaze 1.1 LiteOpen SourceInterfaze's open-weight model for OCR, transcription and extraction, with confidence scores.
Questions about GLM-OCR
What is GLM-OCR?
Parse PDFs, tables, formulas and handwriting into JSON.
Who is GLM-OCR for?
GLM-OCR is best for developers who turn scans and PDFs into clean Markdown or JSON.
Is GLM-OCR free?
Yes. GLM-OCR has a free Self-hosted plan: download from Hugging Face or ModelScope.
How much does GLM-OCR cost?
GLM-OCR's paid plans are Z.ai API at $0.03 per million input or output tokens.
Is GLM-OCR open source?
Yes. GLM-OCR's code is public under the Apache 2.0 licence, so you can run it on your own server.
What are the best GLM-OCR alternatives?
The closest alternatives on Recaply are Cohere Parse 5, Mistral OCR 4.1, Monkt and Interfaze 1.1 Lite.
Where can I try GLM-OCR?
On its official website, docs.z.ai.
Can I run GLM-OCR locally?
Yes. GLM-OCR's weights are on Hugging Face and ModelScope under the MIT License, so you can run it free on your own hardware. Z.ai documents setups for vLLM and SGLang servers, Ollama, and MLX on Apple Silicon Macs, plus a Python SDK installed with pip.
What files and languages does GLM-OCR support?
GLM-OCR reads PDFs and JPG or PNG images; through the API a PDF can be up to 50 MB and 100 pages and an image up to 10 MB. It handles Chinese, English, French, Spanish, Russian, German, Japanese, Korean and other languages.
Google Sheets AI
PDF Dino
Otio
Locus
Genei
Mathesar