Chinese AI Models Capture 30%+ of U.S. Developer Token Usage as OpenAI, Anthropic Costs Rise

Market Share Surge

Chinese AI models have captured over 30% of weekly token usage by U.S. companies on OpenRouter since February 8, 2025, with usage peaking at 46% in some weeks. This represents a dramatic increase from 4.5% in the first half of 2025 and an average of 11% over the previous 12 months, according to data from the AI model routing platform.

OpenRouter enables developers to access multiple AI models through a single interface, making its usage data a key indicator of model adoption patterns among technical teams.

Cost Differential Drives Adoption

Open source Chinese models are "60% to 90% cheaper" than leading models from Anthropic and OpenAI, according to Justin Summerville, who works on data and analytics at OpenRouter. This pricing advantage is reshaping how companies deploy AI.

AI startup Lindy moved 100% of its traffic from Anthropic's Claude models to DeepSeek in June 2025. "We did it, and you could see that cost curve go down, like, crash to the ground," CEO Flo Crivello told CNBC. The company claims the decision will save millions of dollars within months.

Z.ai's GLM 5.2, released in June 2025, saw the fastest adoption of any model tracked by Vercel in 2026. In its first full week after launch, daily token volume grew approximately 27x and customer count grew about 80x, according to Harpreet Arora, head of agentic infrastructure at Vercel.

Performance Gap Narrows

Chinese models are estimated to be "six to nine months" behind top U.S. frontier models in performance, according to Kyle Chan, fellow at the Brookings Institution's John L. Thornton China Center. GLM 5.2 scored within one percentage point of Anthropic's Opus 4.8 on an agentic benchmark while costing roughly one-fifth as much.

DeepSeek's April 2025 model release increased performance on many core use cases for Lindy, according to Crivello. "The new open source models are performing well and prove capable for all but the most complex LLM tasks," Summerville said.

Enterprise Adoption Patterns

GLM 5.2 now ranks in the top five models on LaunchLemonade, an AI agent platform for regulated industries, though Claude and ChatGPT still dominate overall usage, according to CEO Cien Solon. "Chinese models like Z.ai and [Alibaba's] Qwen are becoming options for companies [as] they offer an attractive combination of performance and cost for specific workloads," Solon said.

The shift reflects a broader strategy change among U.S. companies. "Where previously U.S. companies were prioritizing AI adoption regardless of model, now they're getting more cost-conscious," Chan said.

Regulatory Context

The trend emerges as the U.S. administration considers regulation of powerful AI models and foreign alternatives. In late June 2025, OpenAI limited rollout of new models at government request. Export controls on Anthropic's Mythos and Fable models were lifted that month after a standoff between the Trump administration and the company.

What This Means

The rapid adoption of Chinese AI models represents a fundamental market shift driven by cost economics rather than pure performance. With price differentials of 60-90% and performance gaps narrowing to 6-9 months, U.S. companies are increasingly willing to route workloads to cheaper alternatives for tasks that don't require cutting-edge capabilities. This creates strategic pressure on OpenAI and Anthropic to either reduce pricing or further differentiate on capabilities, while raising questions about long-term competitiveness if cost-performance trends continue favoring open-weight Chinese models.