Chinese-built AI models are gaining traction among U.S. businesses as improving performance and significantly lower costs prompt companies to rethink their reliance on leading American systems.
Models from developers including DeepSeek, Z.ai and Alibaba are increasingly being viewed as viable alternatives to offerings from OpenAI and Anthropic, particularly for tasks where cutting-edge performance is not essential.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
Open-Source Models Gain Ground
According to OpenRouter, a platform that provides developers with access to multiple AI models, Chinese models have accounted for more than 30% of tokens used by U.S. companies each week since early February, with their share peaking at 46%.
That marks a sharp increase from an average of 11% over the previous year, reflecting growing demand for open-source and open-weight models as businesses become more focused on controlling AI costs.
The shift comes as Washington tightens oversight of advanced AI technologies. In June, OpenAI delayed the release of a new model at the request of the U.S. government, while export restrictions on Anthropic’s Mythos and Fable models were later lifted.
“Chinese AI models are particularly attractive to American companies now as AI costs skyrocket,” Kyle Chan, a fellow at the Brookings Institution’s John L. Thornton China Center, told CNBC. “Companies are becoming much more cost-conscious.”
Cost Is Driving Adoption
As enterprises scale AI deployments, many are shifting routine workloads to lower-cost models that deliver comparable performance. Unlike most proprietary systems from OpenAI, Anthropic and Google, open-source and open-weight models allow developers greater control over deployment, customization and operating costs. AI startup Lindy recently migrated all of its workloads from Anthropic’s Claude models to DeepSeek.
“We did it, and you could see that cost curve go down, like, crash to the ground,” CEO Flo Crivello told CNBC, adding that the move is expected to save the company millions of dollars.
Vercel also reported rapid adoption of Z.ai’s GLM 5.2 model after its June launch.
“Price is doing the work here,” said Harpreet Arora, the company’s head of agentic infrastructure. “When a task doesn’t need the best model, teams are beginning to route it to the cheapest one that’s good enough.”
According to OpenRouter, some Chinese open-source models can cost between 60% and 90% less than comparable offerings from OpenAI and Anthropic.
Narrowing The Performance Gap
Lower prices are only part of the story. Chinese AI models are also closing the performance gap with leading U.S. systems.
Chan estimates they now trail the most advanced American models by roughly six to nine months, while OpenRouter’s Justin Summerville said the latest open-source models perform well enough for all but the most demanding AI workloads.
GLM 5.2 came within one percentage point of Anthropic’s Opus 4.8 on a widely followed agentic benchmark while costing roughly one-fifth as much. Lindy also reported performance improvements after switching to DeepSeek V4.
“Chinese models like Z.ai and Alibaba’s Qwen are becoming attractive options because they offer a compelling balance of performance and cost,” said Cien Solon, founder and CEO of LaunchLemonade.
As AI adoption accelerates, companies are increasingly weighing whether the highest-performing proprietary models justify their premium pricing. For many everyday workloads, a growing number are concluding that lower-cost Chinese alternatives are good enough.
