Chinese AI Models Are 7.5x Cheaper: Is the US Lead Already Fading?
A new benchmark comparison shows Chinese open-source models crushing US rivals on price, and for most businesses, price is the whole game
A developer ran the same coding task through Anthropic's Claude Opus and China's open-source GLM model. Both finished in about 5.5 minutes. Claude cost $2.33. GLM cost $0.31. That's a 7.5x price difference for functionally identical output, and Andrei Jikh's recent video on this is lighting up YouTube comment sections for good reason.
Yes, the best US model scores 60 on the Artificial Analysis Intelligence Index versus GLM's 51. But here's the inconvenient math: Chinese models including DeepSeek, Qwen, and MiniMax dominate the middle rankings, which is exactly where 90% of real business use cases live. Performance perfectionists can keep paying the Anthropic premium. Everyone else is doing the math.
The US AI narrative has been built on a "best model wins" assumption. China is quietly betting that "cheapest good-enough model wins" and, historically, that's the bet that actually clears the table.
Want the picks behind these posts?
Three AI models grade every call against the S&P 500 — wins and misses published. Free forever.
Create a free account