Chinese AI Models Are Winning Customers On Price. Now OpenAI And Anthropic Face A New Cost War.

As companies expand the use of artificial intelligence across software development, customer service and business operations, many are reassessing which models they use as computing costs continue to climb. That shift is creating new opportunities for Chinese AI developers whose latest models are closing the performance gap with leading American rivals while offering substantially lower prices.

Chinese-built AI models are increasingly being adopted by U.S. companies as businesses look to reduce AI spending without sacrificing performance, CNBC reported Tuesday. The report said models developed by companies including DeepSeek and Z.ai have become attractive alternatives to offerings from OpenAI and Anthropic as enterprises become more cost-conscious.

Data from OpenRouter, a platform that gives developers access to multiple AI models, shows the trend accelerating. CNBC reported that Chinese models have accounted for more than 30% of tokens processed through the platform every week since Feb. 8, with usage climbing as high as 46%. That compares with an average of 11% over the previous 12 months and just 4.5% during the first half of 2025.

Kyle Chan, a fellow at the Brookings Institution’s John L. Thornton China Center, told CNBC that companies initially prioritized AI adoption regardless of which model they used but are now paying much closer attention to operating costs as AI deployments scale.

Developers are also experimenting more frequently with open-source and open-weight models, which allow users to inspect and, in some cases, modify parts of the software. Those models differ from proprietary systems offered by companies such as OpenAI, Anthropic and Google, whose underlying technologies remain closed.

The cost difference has become significant for some businesses. AI startup Lindy recently shifted all of its AI traffic from Anthropic’s Claude models to DeepSeek. Chief Executive Flo Crivello said the move sharply reduced operating costs and is expected to save the company millions of dollars within months.

The report also highlighted strong adoption of Z.ai’s GLM 5.2 model. Harpreet Arora, head of agentic infrastructure at Vercel, told CNBC that GLM 5.2 became the fastest-growing model on the company’s platform this year, with daily token volume increasing about 27 times during its first full week after launch while customer adoption rose roughly 80%.

Justin Summerville, who works on data and analytics at OpenRouter, told CNBC that leading Chinese open-source models can cost between 60% and 90% less than comparable models from Anthropic and OpenAI. He added that the latest Chinese releases are capable of handling all but the most demanding large language model workloads.

The growing adoption reflects improvements in model performance as well as pricing. GLM 5.2 scored within a percentage point of Anthropic’s Opus 4.8 on one widely followed benchmark for AI agents while operating at roughly one-fifth of the cost. Some researchers have also found the model performs competitively on certain cybersecurity benchmarks.

LaunchLemonade, which develops AI agents for regulated industries, has also seen broader adoption of Chinese models. Founder and Chief Executive Cien Solon said that while Claude and ChatGPT remain the platform’s most-used models, GLM 5.2 has already entered its top five, while Alibaba’s Qwen has become an increasingly viable option for businesses seeking lower-cost deployments.

The shift comes as U.S. AI companies navigate tighter government oversight alongside growing competition. Reuters reported last week that the Trump administration lifted export restrictions on Anthropic’s Mythos and Fable models after weeks of negotiations over access to advanced AI systems. The report said federal agencies have simultaneously expanded the use of frontier AI models for cybersecurity work, highlighting the strategic importance Washington places on advanced AI capabilities.

Competition has also intensified around access to enterprise customers. Reuters separately reported that Anthropic has confidentially filed for a U.S. initial public offering while continuing to deepen its work with government agencies and commercial partners. The company has also expanded its enterprise ecosystem through new partnerships, including OZ Digital joining the Anthropic Partner Network to help businesses deploy Claude through Microsoft Azure AI Foundry, according to PR Newswire.

Meanwhile, competition with Chinese developers extends beyond pricing. CNBC reported Monday that Alibaba has prohibited employees from using Anthropic’s Claude Code following concerns surrounding hidden tracking features and Anthropic’s accusations that the Chinese technology company attempted to distill its AI capabilities. Alibaba has not publicly responded to those allegations.

Industry observers say companies are increasingly choosing different AI models depending on the task rather than relying on a single provider. Yacine Jernite, head of machine learning at Hugging Face, told CNBC that many organizations now want AI systems they can control and adapt themselves, making lower-cost open-source and open-weight models an attractive option as enterprise AI adoption continues to expand.