Awesome MLLM Benchmarks

An interactive explorer & paper list for MLLM benchmarks.

🌐 Explore the Interactive Dashboard »
🤗 Sample Data on HuggingFace »

---

We collect 130+ MLLM benchmarks across 20+ categories (OCR, Math, Video, Agent, Spatial Intelligence, etc.) with rich metadata including GitHub stars, HuggingFace downloads, paper abstracts, model evaluation coverage, and more. Everything is presented through a modern interactive dashboard — you can search, filter, sort, and browse actual benchmark samples directly in the browser.

---

Table of Contents

• Getting Started
• Benchmark List
  • OCR / Doc / Chart
  • Spatial Intelligence
  • Knowledge
  • Math
  • Reasoning & Logic
  • STEM
  • Perception
  • Grounding
  • Coding
  • Medical
  • Video
  • Generation
  • Agent
• Contributing
• Citation

---

🚀 Getting Started

Clone and serve locally:

git clone https://github.com/lchen1019/awesome-mllm-benchmarks.git
cd awesome-mllm-benchmarks
python serve.py 8080

Then open http://localhost:8080 in your browser.

serve.py starts a simple HTTP server with no-cache headers. You can also use any other static file server.

Sample images live in samples/ and are served locally with the site. A copy also remains on HuggingFace.

Project Structure

awesome-mllm-benchmarks/
├── index.html                   # Main dashboard page
├── cases.html                   # Sample viewer page
├── data/
│   ├── benchmarks.json          # Benchmark metadata (130+ entries)
│   ├── model-matrix.json        # Model × Benchmark coverage data
│   └── cases-manifest.json      # Sample viewer config
├── samples/                     # Benchmark sample data (images, per-benchmark data)
├── assets/                      # Logo, favicon, screenshots
├── serve.py                     # Dev server with no-cache headers
├── .github/workflows/pages.yml  # GitHub Pages deployment
└── .nojekyll                    # Bypass Jekyll processing

---

📚 Benchmark List

OCR/Doc/Chart

Spatial Intelligence

Knowledge

Math

Reasoning & Logic

STEM

Perception

Grounding

Coding

Medical

Video

Video / Knowledge

Video / Reasoning

Video / Motion

Video / Long-form

Video / Streaming

Video / Multi-task & Multiple

Generation

Structured Output & Visualization

Agent

Agent / General

Agent / Computer

Agent / Browser

Agent / Mobile

Agent / Visual Search

---

🤝 Contributing

Contributions are welcome — whether it's adding new benchmarks & data, improving visualizations, or fixing errors. Feel free to open a PR!

---

📝 Citation

If you find this project useful for your research, please consider citing:

@misc{awesome-mllm-benchmarks,
  title        = {Awesome MLLM Benchmarks: An Interactive Explorer for MLLM Benchmarks},
  author       = {Chen, Lin},
  year         = {2026},
  howpublished = {\url{https://github.com/lchen1019/awesome-mllm-benchmarks}},
  note         = {Accessed: 2026}
}