China’s open weight AI push narrows gap with US, report says

From the vantage point of global politics, the most important detail in this story is not any single Chinese model benchmark but the narrative Beijing is building around them. By putting open weight models like DeepSeek, GLM 5.2 and Kimi K3 at the centre of its message, China is positioning itself as the champion of “inclusive” and “open” AI against a more closed US stack. At the same time, technical experts quoted in the piece argue that, for many practical tasks, top Chinese models are now close enough to frontier US systems that price and deployment flexibility matter more than a few points on a leaderboard.

That combination is potent in the race to AGI because it blends capability with coalition building. If Chinese open weight models can be cheaply fine tuned and self hosted in countries across the Global South, they become an attractive alternative to relying on US based APIs that may be constrained by export controls or political conditions. The launch of a China led World AI Cooperation Organisation to counter a US backed “Pax Silica” illustrates how standards, infrastructure finance and ideological framing are all being weaponised. Even if US labs stay slightly ahead on absolute frontier performance, the diffusion and normalisation of near frontier Chinese models could matter just as much for where de facto AI power sits.