Chinese AI models gain traction among US companies due to cost benefits

Data from OpenRouter, the world's largest AI aggregation platform, shows that since February 8, Chinese AI models have been increasingly adopted by US companies. Weekly usage of models like DeepSeek and Zhipu's GLM series among US firms has stayed above 30%, peaking at 46%, compared to an average of 11% over the past 12 months.

AI startup Lindy switched all its traffic from Anthropic's Claude to DeepSeek in June, with its CEO saying the move will save millions of dollars over months. OpenRouter's data chief noted Chinese open-source models can be 60% to 90% cheaper than leading models from Anthropic and OpenAI.

The performance gap is closing too. Brookings' Chan estimates Chinese models trail top US ones by just 6 to 9 months. Benchmarks show GLM 5.2 trails Anthropic Opus 4.8 by only 1 percentage point on an Agent benchmark, while costing about one-fifth as much.