Open Weights, Closed Frontiers: Small Language Models, Data Sovereignty & the Enterprise Model-Portfolio Decision

Kim, Kibaek

The paper weighs the competitive advantages of both open and closed ecosystems across capability, data sovereignty, and licensing and adds a specific angle: that aggregating several fine-tuned small open models into a private "collective intelligence" ensemble can rival a single large model while keeping data on-premise and the costs low, using software quality-assurance orchestration as the applied example.


More About This Work

Expert Reviewer: Steve Rothman. View our expert review panel →

Author Affiliation: '25SPS, Applied Analytics