A Cartography of Open Collaboration in Open Source AI: Mapping Practices, Motivations, and Governance in 14 Open Large Language Model Projects

Researchers conducted an in-depth study of 14 open-source large language model projects through developer interviews, revealing how collaboration, governance, and participation evolve across different development stages. The study maps motivations ranging from democratizing AI to expanding language representation, showing that openness in open-source AI emerges from complex interactions between artifact domains, lifecycle stages, and institutional contexts rather than being a uniform property.

This empirical study provides the first systematic examination of how open LLM projects actually operate beyond their public release, addressing a significant gap in understanding the rapidly expanding ecosystem of open artificial intelligence. Rather than treating open-source AI as monolithic, the researchers identify how collaboration patterns shift dramatically from concentrated, selective engagement during early development to distributed, broader participation after public release. This distinction matters because it reveals that successful open LLM ecosystems require different governance structures and participation strategies depending on development stage.

The research contextualizes a broader trend toward decentralization in AI development, driven partly by frustration with proprietary models and partly by genuine technical benefits of collaborative approaches. Open LLM projects now compete directly with commercial models from major tech companies, creating incentives for developers to organize effective collaboration frameworks. The study identifies multiple motivations—democratization, regional ecosystem building, language representation, and open science—suggesting that open LLM development serves diverse stakeholder interests rather than pursuing a single objective.

For the AI industry, this research validates that open-source development at scale requires professional governance structures alongside grassroots contributions. Organizations investing in open LLM infrastructure can learn which coordination mechanisms succeed across different artifact domains—models, data, software, evaluation, compute, and community engagement. This framework helps developers and institutions navigate tradeoffs between openness and control. The findings suggest that sustainable open LLM ecosystems balance centralized stewardship with distributed participation, adapting structures as projects mature and communities expand.

• →Open LLM collaboration patterns fundamentally shift from concentrated early development to distributed post-release participation
• →Successful open-source AI requires multi-domain coordination across models, data, software, evaluation, compute, and community engagement
• →Developer motivations span democratization, regional ecosystems, language diversity, and open science rather than single objectives
• →Governance structures range from centralized company-led efforts to decentralized grassroots initiatives, each requiring different participation strategies
• →Openness in open-source AI emerges from interconnected organizational choices rather than being an inherent or uniform property