The biggest shift in AI right now isn't a new model, it's who's winning. Chinese open-weight models (DeepSeek, Kimi, MiniMax, GLM) now make up 61% of all tokens used on OpenRouter. US models fell from 70% to 30% share in a single year. Why? They're 10-20x cheaper, you can self-host them, and they now match or beat top US models on real coding benchmarks. Stay ahead at www.contentbuffer.com Have you tried an open-weight model yet? #AI #LLM #OpenSource #DeepSeek #TechNews #MachineLearning