🌳 ModelForest

Interactive visualization of AI model genealogy — see how every major LLM, image model, and embedding model evolved from its ancestors.

🔗 Live Demo →

---

What is this?

ModelForest is a free, open-source, interactive family tree of AI models. Like ancestry.com for AI models — a zoomable, searchable, filterable tree showing:

• 🔗 Parent → Child relationships (base model → fine-tune)
• 🏗️ Architecture lineages (Transformer → BERT → RoBERTa → DeBERTa)
• 🏢 Company timelines (OpenAI, Meta, Google, Anthropic, Mistral...)
• 🔀 Model merges and distillations
• ⭐ Key milestones and breakthrough moments

Model Families Included

Family	Company	Models
GPT-1 → GPT-4.1 & o1-o4	OpenAI	19
LLaMA → Llama 4	Meta	19
Transformer → BERT → T5 & PaLM → Gemini 2.5	Google	25
Claude 1 → Claude 4	Anthropic	12
Mistral → Mixtral & Codestral	Mistral AI	12
DeepSeek → R1	DeepSeek	8
Phi-1 → Phi-4	Microsoft	9
Qwen → Qwen 3	Alibaba	7
Stable Diffusion → FLUX	Stability AI / BFL	10

Features

• 🌳 D3.js Collapsible Tree — expand/collapse model branches
• 🔍 Fuzzy Search — find any model instantly (Ctrl+K)
• 🎨 Color-coded by company
• 📊 Detail Sidebar — click any model for full info
• 🎛️ Filters — by company, modality, license, year
• 🌐 Radial View — alternative circular layout
• 📱 Responsive — works on desktop, tablet, mobile
• 🤖 Auto-updating — daily HuggingFace API scrape via GitHub Actions
• 🆓 Free forever — static site on GitHub Pages

Tech Stack

Layer	Technology
Visualization	D3.js v7
Search	Fuse.js
Styling	Custom CSS (dark theme)
Fonts	Inter + JetBrains Mono
Data Pipeline	Python + huggingface_hub
Automation	GitHub Actions (daily cron)
Hosting	GitHub Pages

Project Structure

ModelForest/
├── .github/workflows/     # CI/CD & auto-scraping
├── data/
│   ├── curated/           # Hand-verified model data (9 family files)
│   ├── scraped/           # Auto-scraped from HuggingFace API
│   ├── companies.json     # Company metadata & colors
│   └── stats.json         # Summary statistics
├── scripts/
│   ├── scrape_hf.py       # HuggingFace API scraper
│   ├── build_tree.py      # Data merge & tree builder
│   └── validate_data.py   # Data integrity checks
└── src/
    ├── index.html          # Main page
    ├── css/style.css       # Styles
    └── js/
        ├── app.js          # Main entry point
        ├── tree-view.js    # D3 tree visualization
        ├── search.js       # Fuzzy search
        ├── sidebar.js      # Detail panel
        └── filters.js      # Filter controls

Development

Local Setup

# Clone
git clone git@github.com:MrUnreal/ModelForest.git
cd ModelForest

# Serve locally (any static server works)
npx serve src
# or
python -m http.server 8000 -d src

Open http://localhost:8000 — the app loads data from ../data/ relative to src/.

Run Data Pipeline

# Install Python deps
pip install -r scripts/requirements.txt

# Validate curated data
python scripts/validate_data.py

# Build merged data files
python scripts/build_tree.py

# Scrape HuggingFace (optional, needs HF_TOKEN)
HF_TOKEN=your_token python scripts/scrape_hf.py

Contributing

Add a Model

1. Open an issue using the "Add Model" template
2. Or submit a PR adding the model to the appropriate data/curated/*.json file

Model Data Format

{
  "id": "org/model-name",
  "name": "Model Name",
  "shortName": "Short Name",
  "company": "Company",
  "releaseDate": "2024-01-01",
  "parameters": "7B",
  "parametersRaw": 7000000000,
  "architecture": "LlamaForCausalLM",
  "architectureFamily": "Transformer (Decoder-only)",
  "modality": "text",
  "license": "apache-2.0",
  "licenseType": "open-source",
  "contextWindow": 8192,
  "description": "Brief description.",
  "source": "curated"
}

Relationship Types

Type	Description
base	Direct next version
finetune	Instruction-tuned variant
distillation	Smaller student model
merge	Multiple models merged
architecture	Architectural derivative
quantization	Quantized variant
multimodal-extension	Added new modality

License

MIT — see LICENSE

---

Built by MrUnreal • Data from HuggingFace