China's cheap, open AI models are undercutting top American ones, which is good for businesses watching costs and bad for US AI companies counting on premium pricing.

• Chinese models like GLM, DeepSeek, and Qwen do the same work as top US models for a fraction of the cost — one coding test came in about 7 times cheaper.
• The US still has the smartest AI, scoring highest on the main global ranking, but Chinese models fill the entire middle of the pack.
• For most business tasks — customer service emails and routine "boring" work — companies do not need the smartest model, just a good-enough cheap one.
• That gap makes China's open models the practical winner for the 90% of AI work that is ordinary business grunt work.

Outlook: Expect more businesses to pick cheap Chinese models for everyday tasks, squeezing the pricing power of top US AI firms.