Chinese AI models seize up to 46% of US developer use

TL;DR:

• Chinese models’ share of tokens routed by US firms through OpenRouter has held above 30% since February, peaking at 46%, against an 11% average over the previous year.
• AI startup Lindy moved 100% of its traffic from Claude to DeepSeek, a switch it says will save millions; on Vercel, Z.ai’s GLM-5.2 saw daily token volume grow 27-fold in its first week.
• Open Chinese models can run 60% to 90% cheaper than leading US systems.

The headline this week is not another benchmark result but hard adoption data. As token prices rise at US labs, engineers are routing more work to cheaper open-weight models — and the most capable of those increasingly come from China. “Price is doing the work here,” Vercel’s Harpreet Arora told CNBC. “When a task doesn’t need the best model, teams are beginning to route it to the cheapest one that’s good enough.”

From capability to procurement

This builds on a trend Resultsense has tracked as GLM-5.2 closed the performance gap with US leaders and soaring bills pushed businesses towards cheaper models. What is new is the scale of switching now visible in the routing data. GLM-5.2 landed within a percentage point of Claude Opus 4.8 on one agentic benchmark at roughly a fifth of the cost; Brookings estimates Chinese models trail top US rivals by just six to nine months.

Looking forward

For UK enterprises, the calculus is sharpening. Hugging Face’s Yacine Jernite warns of a real risk that users get “stuck” choosing between expensive US proprietary models with volatile pricing and Chinese models as the only affordable way to own their stack. That is not a comfortable choice for firms in regulated sectors weighing data-security concerns against runaway inference costs — and it makes model procurement, not model capability, the decision that increasingly matters.