Tiny, cheap and good at tables and formulas, so document pipeline builders should test it before paying for bigger OCR models.
Kenning's take
What it does
GLM-OCR is a 0.9B-parameter model from Z.ai for reading documents and images, including tables, formulas and forms. Z.ai says it topped the OmniDocBench V1.5 benchmark at launch. It runs through the Z.ai layout parsing API or locally from open weights.
- 0.9B parameters
- Text, table and formula recognition
- Layout parsing of PDFs and images
- Structured extraction from forms and invoices
- Open weights for local use
Who it suits. Developers building document processing or data extraction pipelines.
Key facts
- Category
- AI models
- Made by
- Z.ai (Zhipu AI)
- Based in
- China
- Launched
- 2026
- How to get it
- Z.ai layout parsing API or self-hosted from open weights
- Works on
- API, open weights (Hugging Face)
- In the UK
- Yes
- Free plan
- weights free to download
- API price
- $0.03 per million tokens, input and output
Reviews and write-ups
- NYU Shanghai RITS 2026-02-24
FAQ
Is GLM-OCR open source?
Yes, the weights are on Hugging Face and can be run locally.
How much does the GLM-OCR API cost?
$0.03 per million tokens for both input and output.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.