Open-source Chinese LLMs challenge proprietary models, shifting value to hardware

A discussion highlights how rapid advancements in open-weight Chinese LLMs, such as Qwen and GLM, are closing the performance gap with top-tier proprietary models. This trend threatens the business models of proprietary AI companies as organizations can host these highly capable open-source alternatives themselves. If open-source models achieve parity with proprietary giants, the economic value of AI may shift away from companies selling API tokens to hardware manufacturers like Nvidia and AMD that supply the necessary compute. This could force proprietary AI developers to rethink their monetization strategies. The post references spending data showing that businesses spend far less on ultra-premium models compared to mid-tier ones, suggesting that highly capable open-source models can easily capture the bulk of the market. However, the analysis uses confusing terminology (like 'Fable 5' and 'Opus 4.8') to describe proprietary model tiers.