Dolphin (ByteDance) vs LangExtract (Google)

Dolphin (ByteDance) is free, with no paid plan attached. LangExtract (Google) is free, with no paid plan attached. Both are listed under AI OCR & Document Extraction, so this is a like-for-like comparison. Neither is ranked above the other — Flocci carries no sponsored placement.

Dolphin (ByteDance) vs LangExtract (Google) — straight answers

Dolphin (ByteDance) vs LangExtract (Google): what is the difference?

Dolphin (ByteDance) is free, with no paid plan attached and is listed for two-stage document-type and layout analysis then element parsing. LangExtract (Google) is free, with no paid plan attached and is listed for python library extracting structured data from unstructured text with LLMs. Both sit in AI OCR & Document Extraction.

Is Dolphin (ByteDance) or LangExtract (Google) cheaper to start with?

Neither — Dolphin (ByteDance) and LangExtract (Google) are both free, so the choice comes down to capability rather than cost. Both entries list what they uniquely do above.

Which should I choose, Dolphin (ByteDance) or LangExtract (Google)?

Choose Dolphin (ByteDance) if you need two-stage document-type and layout analysis then element parsing; choose LangExtract (Google) if you need python library extracting structured data from unstructured text with LLMs. Flocci AI Tools does not rank one above the other — it shows both feature sets side by side and lets the requirement decide.

Dolphin (ByteDance) compared with LangExtract (Google): pricing tier, category, listed capabilities and links.
 Dolphin (ByteDance)LangExtract (Google)
Pricing tierFreeFree
Free to startYesYes
CategoryResearch & LearningResearch & Learning
TypeAI OCR & Document ExtractionAI OCR & Document Extraction
Listed capabilities
  • Two-stage document-type and layout analysis then element parsing
  • Page-level and element-level parsing modes
  • Hugging Face Transformers integration
  • MIT licensed; Dolphin-v2 released Dec 2025
  • Python library extracting structured data from unstructured text with LLMs
  • Maps each extraction to its exact source span
  • Works with Gemini, OpenAI and local Ollama models
  • Interactive HTML visualization of results
Tagsdolphin ocr, bytedance dolphin, dolphin-v2 document parsing, open source document parser, free document image parsinglangextract, google langextract, extract structured data from text llm, source grounded extraction, gemini extraction library
Websitegithub.comgithub.com
Full pageDolphin (ByteDance) details →LangExtract (Google) details →
AlternativesDolphin (ByteDance) alternatives →LangExtract (Google) alternatives →

DeepSeek-OCR

AI OCR & Document Extraction
freeNew
  • Vision-language OCR that compresses document context optically
  • Document-to-Markdown, layout detection and figure parsing
  • Multiple resolution modes up to 1280x1280
  • MIT license; DeepSeek-OCR2 released Jan 2026

Docling

AI OCR & Document Extraction
free
  • MIT-licensed, 64.7k GitHub stars, IBM Research-originated, now under LF AI & Data Foundation
  • Advanced PDF layout + table structure recognition, local execution
  • Exports to Markdown/HTML/JSON, integrates with LangChain/LlamaIndex

dots.ocr (dots.mocr)

AI OCR & Document Extraction
freeNew
  • 3B vision-language model for layout parsing across many scripts
  • Converts charts and diagrams into SVG code
  • Scores 83.9 on olmOCR-bench
  • MIT licensed with free live demo

Marker

AI OCR & Document Extraction
free
  • Apache-2.0 code, 38.7k GitHub stars, converts PDFs/DOCX/PPTX/XLSX to markdown/JSON
  • 76.0% accuracy on the olmocr-bench benchmark
  • Runs on GPU, CPU, or Apple Silicon; optional LLM-boosted accuracy mode

MarkItDown

AI OCR & Document Extraction
freeNew
  • Converts PDF, Office files, images, audio and HTML into LLM-friendly Markdown
  • Ships an MCP server for agents
  • Open-source Python package and CLI
  • Optional LLM-based image description

MinerU

AI OCR & Document Extraction
freeNew
  • Parses PDFs, images and Office files into Markdown, JSON, HTML or LaTeX
  • Runs fully locally with OCR, table and formula recognition
  • Free hosted web app at mineru.net plus agent-ready outputs
  • Version 4.0 adds four parsing quality tiers