Author(s): Caspar Bannink – AI Engineer

Originally published on Towards AI.

Closed Source, Open Source Or Open Weights, Which Is Winning The AI Race?

The AI race has split into two races.

After the introduction, the article argues that “closed models,” “open weights,” and “fully open systems” are distinct strategies rather than interchangeable labels. It defines what “open” really should mean, contrasting Open Source AI (open parameters plus training/run code and detailed data information) with open weights (only downloadable parameters) and closed source (managed access via app/API). It then explains why closed models often win the frontier capability race—because they centralize expensive compute and feedback—while open weights win distribution by traveling through many deployment environments and tool stacks. The piece further shows how licensing shapes what “self-host” truly enables, warning that terms like Apache/MIT versus custom licenses can determine whether builders can build sustainable businesses. Finally, it contrasts fully open systems as a win for scientific credibility and reproducibility, concludes that no single approach dominates all dimensions, and suggests that the strongest companies will likely combine all three loops (managed performance, distributed adoption, and transparent research) while aligning strategy to which parts of the stack they want to own.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI