Chinese open-weight models and domestic AI platforms compete for ecosystem influence

Z.ai’s GLM-4.5, Qwen, DeepSeek, Tencent’s reported Hy3, Moonshot AI’s Kimi K3, and other efforts are competing for developer adoption, international use, and technology partnerships. ModelScope, MoArk, and related platforms are building China-centered repositories and tooling, but developers’ continued preference for Hugging Face, uncertain openness and cross-border access, and limited verified industrial uptake temper claims of global leadership; Anthropic’s model-diversion allegations remain unverified.