This is the best summer for Chinese AI: GLM 5.2 -> KIMI K3 -> Qwen 3.8 Max. Only Minimax hasn’t turned in their paper yet—if they don’t deliver, it might really break down. A lot of people think China’s leading AI models are all just distilled versions of Anthropic. If that’s true, then why don’t other countries have powerful open-source models? And why haven’t they been blocked by compute power limitations? The truth is, the original innovations in China’s open-source models go far beyond what distillation can explain. Isn’t storage always a bottleneck? Why not let Korea try distillation? India has the world’s top software engineers—why haven’t they made something comparable and instead become the ones doing the distilling? It’s because China is the only country outside of Silicon Valley with a complete AI industry chain: hardware manufacturing, chip design, application scenarios, and internet ecosystems.