A challenge to assumptions about Chinese AI model prices and US company spending

A post argues that two common AI cost assumptions are outdated. It cites @ArtificialAnlys for a finding that US models cost less per task at almost every intelligence level, even if Chinese models might cost less per token. Citing @arakharazian, it says US companies' AI use is rising while costs are leveling off, alongside better cost-capability choices and deals driven by competition between Anthropic and OpenAI.