Late open-source LLM derivatives face weak community recognition

The open-source AI community might celebrate collaboration, but this paper proves it's still a zero-sum game for attention. Being first matters more than being slightly better.

• –Derivative models (fine-tunes, quants) targeting prominent base models face a rapid "race to the bottom" for visibility
• –Late entries to a crowded model category receive significantly less community recognition
• –Platform feedback loops on Hugging Face heavily reward early movers over latecomers
• –Suggests developers should prioritize speed or highly differentiated niches over incremental improvements on saturated base models