Z.ai - Inspiring AGI to Benefit Humanity

GLM-OCR is a lightweight professional OCR model with parameters as small as 0.9B, yet it achieves state-of-the-art performance across multiple capabilities. It sets a new benchmark for document parsing with its “small size and high accuracy.” Key features include:

Performance SOTA: Scored 94.62 points to top OmniDocBench V1.5 and achieved current best performance across multiple mainstream document understanding benchmarks including tables and formulas at launch.
Optimized for Real-World Scenarios: Delivers stable, leading accuracy in complex environments like code documentation, intricate tables, and stamp recognition. Maintains exceptional recognition precision even with complex layouts, diverse fonts, or mixed text-image content.
Efficient and Cost-Effective: With just 0.9B parameters, supports VLLM and SGLang deployment, significantly reducing inference latency and computational overhead.

Input Modality

PDF, images (JPG, PNG)
Single image ≤ 10MB, PDF ≤ 50MB
Maximum support: 100 pages

Output Modality

Text / Image Links / MD Documents

Supported Language

Support Chinese, English, French, Spanish, Russian, German, Japanese, Korean, etc.

Performance

State-of-the-Art Performance, Precision in Action

Thanks to its proprietary CogViT visual encoder and deep scene optimization, GLM-OCR achieves “compact size, high accuracy.”With only 0.9B parameters, GLM-OCR achieved SOTA on the authoritative document parsing benchmark OmniDocBench V1.5 with a score of 94.6. It outperforms multiple specialized OCR models across four key domains—text, formula, table recognition, and information extraction—with performance approaching that of Gemini-3-Pro.

Beyond public benchmarks, we conducted internal evaluations across six core real-world scenarios. Results show GLM-OCR delivers significant advantages across dimensions including code documentation, real-world tables, handwriting, multilingual text, seal recognition, and invoice extraction.

Faster, More Cost-Effective

For speed, we compared different OCR methods under identical hardware and testing conditions (single replica, single concurrency), evaluating their performance in parsing and exporting Markdown files from both image and PDF inputs. Results show GLM-OCR achieves a throughput of 1.86 pages/second for PDF documents and 0.67 images/second for images, significantly outperforming comparable models.