Is anyone here using Chinese models or AI agents?

A developer notes conflicting views on Chinese AI models, with some claiming dominance on platforms like OpenRouter due to cost, while others cite heavy censorship, slow performance, and weak English. The developer's personal experience with hy3 through WorkBuddy, which offers models like GPT-6-Astra and Gemini-3.5-Flash alongside its own, suggests these models handle 80% of their work effectively, making Codex seem like overkill. They are also interested in trying GLM-5.3, Kimi K3, and DeepSeek V4.1 Flash.