AI · AI model
NuMarkdown-8B-Thinking
Open-source vision-language model converting complex documents into structured Markdown using reasoning.
NuMarkdown-8B-Thinking is an open-source vision-language model fine-tuned from Qwen2.5-VL-7B designed to convert document images and PDFs into clean Markdown. It generates internal reasoning tokens to analyze document layout and table structures before producing the final text. This architecture makes it especially effective for ingesting visually complex documents into RAG pipelines.
What it's for
- Convert complex scanned PDFs and document images into structured Markdown
- Extract nested and multi-row tables cleanly for RAG ingestion
- Parse multi-section document layouts without losing hierarchy or text flow
Features
- Generates thinking tokens to analyze layout and merged cells prior to Markdown output
- Fine-tuned on synthetic reasoning traces with GRPO reinforcement learning
- Specialized handling for irregular layouts, multi-column blocks, and complex tables
- Compatible with Hugging Face Transformers and deployable via vLLM
- Released under an open-source MIT license
Pricing
Open source
Details
- Type
- AI model
- Category
- AI
- Topic
- Document Parsing
- Open source
- Yes
- Platforms
- Self-hosted, Code library, API
- For
- AI engineers, RAG developers, and researchers processing complex documents
- Alternative to
- OCRFlux-3B, GPT-4o, Gemini 2.5 Flash
- Tags
- ocr, vlm, markdown, qwen, rag, reasoning
Why it was saved
- @askgpts: Accurately extracts text, tables, and document structures from complex PDFs, scans, and screenshots directly into clean Markdown for RAG pipelines and automation.
Compare
Related
Open 10B multimodal model with parallel reasoning competitive with frontier systems
Open-source vision-language model converting complex document images to structured Markdown
Open-source vision-language model for parsing multi-page documents in one pass
Generates llms.txt and llms-full.txt markdown files from any website URL.
Agentic document parsing and extraction platform for AI pipelines.
Apify web crawler that extracts clean text and Markdown for LLMs
Updated