UV OCR Scripts: 14 Vision-Language Models for Markdown Extraction

14 OCR models ranging from 0.9B to 8B parameters provided by uv-scripts as of March 2026. These scripts facilitate the conversion of image-based datasets into markdown format using HuggingFace Jobs and the UV package manager.

Use Cases

• Automated markdown extraction from document images using glm-ocr.py
• Scaling OCR processing across large datasets using HuggingFace Jobs flavors like l4x1
• Comparing OCR accuracy across 14 different model parameter sizes

Strengths

• Includes 14 distinct OCR models
• Supports model scales from 0.9B to 8B parameters
• Outputs standardized markdown format

Limitations

• Does not contain a static dataset of images
• Requires external input datasets to generate results
• Dependent on HuggingFace Jobs infrastructure for execution

Provenance

Source

uv-scripts

Freshness

Last updated March 2026