[Research] Add 2 elite-tier projects to Open Foundation Models

[alvinreal]

[Research] Add 2 elite-tier projects to Open Foundation Models

Summary

This PR adds 2 battle-tested, production-proven open foundation model projects to the Open Foundation Models category.

Projects Added

1. MMaDA (Gen-Verse) ⭐ 1,620 stars | MIT License

   • Open-sourced multimodal large diffusion language model (dLLM)
   • Unified architecture for text generation, image generation, and multimodal reasoning
   • Key innovations: block diffusion, mixed chain-of-thought fine-tuning, unified RL (UniGRPO)
   • Published at NeurIPS 2025, actively maintained (last update Feb 2026)
   • Category: Multimodal Models (Vision + Language)

2. faster-whisper (SYSTRAN) ⭐ 21,985 stars | MIT License

   • Reimplementation of OpenAI's Whisper using CTranslate2 inference engine
   • Up to 4x faster than original Whisper with same accuracy
   • Supports batched processing and 8-bit quantization for lower memory usage
   • Essential production tool for speech-to-text pipelines
   • Category: Speech & Audio Models

Elite Criteria Verification

Project	Stars	Active (6mo)	Production	Quality	License
MMaDA	1,620 ✓	Feb 2026 ✓	NeurIPS 2025, HuggingFace demo	Full docs, tests	MIT ✓
faster-whisper	21,985 ✓	Nov 2025 ✓	Benchmarked, widely adopted	Full docs, releases	MIT ✓

Validation

• Local validation passed (python3 tools/validate_awesome.py --skip-remote)
• Single category only (Open Foundation Models)
• README.md only changes
• Non-structural (entry additions only)
• OSI-approved licenses

---

Automated research loop run for category: Open Foundation Models