CNBC reported on the 7th that rising pricing from leading U.S. AI vendors is prompting rapid uptake of Chinese large AI models in U.S. enterprise deployments on cost grounds. Industry sources say top Chinese open-source and open-weight models lag frontier U.S. models (OpenAI, Anthropic) by roughly 6–9 months on performance but are priced about 60%–90% lower and cover most routine AI tasks. Aggregator OpenRouter data shows Chinese models have accounted for over 30% of tokens called by U.S. firms each week since Feb. 8, peaking at 46%; the prior 12‑month token-share average was 11%.(金十数据)