ByteDance is not open-sourcing its leading C-Dance 2.5 video model, violating the common narrative of China's open-source push. This reveals a pragmatic strategy: open-sourcing is a tool for challengers to build an ecosystem, but incumbents with distribution advantages will keep their frontier models closed to protect their lead.

Chinese AI models are largely open source not for ideological reasons, but as a pragmatic branding strategy. Open-sourcing their models was necessary to build trust and credibility with Western developers who might otherwise be skeptical of closed, proprietary Chinese technology.

According to SemiAnalysis, multiple major Chinese AI labs are signaling to inference providers that their next frontier models will not be open source. Instead, they will be available only through licensing, suggesting a rapid decline in the open-source movement for top-tier models.

China's push for open-source AI may not be purely strategic but a consequence of US export controls limiting their inference compute. Unable to monetize closed APIs effectively to a skeptical Western market, Chinese labs release models to build influence and attract talent, as few would pay for a sub-frontier, China-hosted service.

Alibaba's release of three proprietary models in three days, with its CEO taking direct control to maximize revenue, marks a decisive shift away from open source. This reflects a broader trend among Chinese tech giants to prioritize direct monetization and commercialization over community-based model development.

Companies like Z.ai are not abandoning open source but using it strategically. They release lightweight models to attract developers and build a user base, while reserving their most powerful, agentic systems for proprietary, revenue-generating enterprise products, creating a clear monetization funnel.

China isn't giving away its AI models out of generosity. By making them open source, it encourages widespread adoption and dependency. Once users are locked into the ecosystem, China can monetize it, introduce ads, or simply lock down future, more advanced versions, giving it significant strategic leverage.

A common misconception is that Chinese AI is fully open-source. The reality is they are often "open-weight," meaning training parameters (weights) are shared, but the underlying code and proprietary datasets are not. This provides a competitive advantage by enabling adoption while maintaining some control.

China's strategy of releasing powerful open-weight models to "fast follow" US capabilities creates a paradox. While it closes the technology gap, it prevents Chinese labs from building sustainable businesses. Without monetizing via proprietary APIs like OpenAI, they cannot generate the revenue needed to acquire the massive compute resources required for long-term competition.