AI Competition Shifts to Open Source vs Closed Source as Costs Drop 60%

Global AI rivalry pivots from US-China tensions to open versus closed source models. DeepSeek V4 Flash trails GPT-5.6 Luna by one point but offers 60% lower costs, driving industry adoption.

Woofun AI notes that the global artificial intelligence competitive landscape has transitioned from a United States-China confrontation to a contest between open-source and closed-source architectures. Data from Artificial Analysis indicates that DeepSeek V4 Flash possesses an intelligence index only one point below GPT-5.6 Luna, while maintaining a single-task cost 60% lower despite an 80% price reduction by OpenAI.

The commentary argues that export controls have inadvertently spurred architectural innovation in Chinese firms, leading to open-weight models that leverage overseas cloud infrastructure for low-cost global deployment. This model mirrors open-source software economics, where revenue derives from hosting rather than self-deployment. Industry figures including David Sacks and Jensen Huang have recently advocated for open-source adoption, with calls for US-China cooperation on AI security to treat the technology as a global public good.