China's AI models undercut US rivals in price as they close performance gap

Chinese AI models now rival top US systems in performance while undercutting them on price and openness, raising questions about whether Beijing can eventually seize global AI leadership. Alibaba's Qwen3.8-Max became China's top-ranked text model, while DeepSeek's V4-Flash costs about three cents per test versus $1.86–$3.15 for leading US rivals. Chinese open models also surpassed American ones in downloads: 663 million versus 595 million. The US retains the edge in frontier models, with $285.9 billion in private AI investment versus China's $12.4 billion and superior chip access, though Beijing is boosting domestic processors. Z.ai's GLM-5.5 releases this month.