In a striking demonstration of shifting dynamics in the global artificial intelligence market, Chinese AI models have outpaced their US counterparts in API call volume for six consecutive weeks. The data, sourced from OpenRouter, one of the world’s largest AI model routing platforms—highlights how aggressive pricing strategies and rapid performance improvements are driving global developers toward Chinese foundational models.
According to a report by Pandaily, during the most recent tracking period, the platform processed 36.1 trillion tokens in global API requests over a single week. Within that volume, models developed by Chinese firms handled 14.19 trillion tokens, establishing a massive lead over the 3.2 trillion tokens routed to models built by American companies.
The Rise of the Chinese “Value” Model
The surge in Chinese API usage is largely attributed to the brutal AI price wars that have defined the domestic market in 2026. By drastically reducing inference costs, Chinese AI companies have made their models highly attractive to global developers who are increasingly prioritizing cost efficiency over raw, bleeding-edge benchmark performance.
Leading the charge is DeepSeek-V4-Flash, which secured the number one position in global API call volume. DeepSeek’s strategy of offering strong reasoning capabilities at a fraction of the cost of Western alternatives has clearly resonated with developers building high-volume applications.
Following closely in the number two spot is Tencent’s Hy3 Preview. Released as an open-source offering, the model is built on a Mixture-of-Experts (MoE) framework featuring 295 billion parameters and an expansive 256K context window, making it a powerful tool for complex, data-heavy tasks.
MiniMax M3 Breaks the Top Three
Perhaps the most notable development in the recent data is the ascent of MiniMax M3, which broke into the global top three for the first time. MiniMax, a Shanghai-based startup, has aggressively positioned its M3 model as a high-performance, cost-effective alternative to established Western models. Its success on OpenRouter indicates that the company’s strategy is gaining traction beyond China’s borders.
Rounding out the strong Chinese showing is Xiaomi’s MiMo-V2.5, which secured the number four position. The dominance of these four Chinese models underscores a broader trend: developers are finding that Chinese models offer comparable performance for the vast majority of commercial use cases, but at a significantly lower price point.
The Decline of Western Dominance
The rise of Chinese models on OpenRouter has come at the direct expense of prominent US models. Notably, Anthropic’s Claude Sonnet 4.6 and Claude Opus 4.7 both dropped out of the top five entirely during this tracking period. OpenAI’s models, while still widely used, have also seen their relative share decline as developers increasingly route requests to lower-cost Chinese alternatives for standard tasks.
This shift suggests that while US companies may still hold a slight edge in absolute frontier capabilities, that advantage is not translating into dominant market share for everyday API usage. For many developers, the marginal performance gains offered by the most expensive US models do not justify the premium pricing, especially when Chinese alternatives like DeepSeek and MiniMax provide “good enough” performance at a fraction of the cost.
The Infrastructure Behind the Numbers
The six-week streak of Chinese dominance on OpenRouter is not simply a product of aggressive pricing. It also reflects the maturation of Chinese AI infrastructure. Chinese cloud providers, including Alibaba Cloud, Tencent Cloud, and Baidu AI Cloud, have invested heavily in building low-latency, globally accessible API endpoints that can serve international developers with competitive response times.
Furthermore, the open-weight nature of many leading Chinese models—particularly those from DeepSeek and Alibaba’s Qwen series—has enabled a global ecosystem of third-party API providers to offer these models at extremely competitive prices. Developers can access these models not only through official channels but through a growing network of independent inference providers who compete aggressively on price and performance, further driving down costs and increasing accessibility.
Implications for the Global AI Race
The OpenRouter data carries significant strategic implications for the global AI race. API call volume is not merely a vanity metric—it represents the actual deployment of AI capabilities in real-world applications. When developers build their products on top of Chinese foundational models, they create deep dependencies that are difficult and costly to reverse. This lock-in effect means that today’s API usage patterns will shape the competitive landscape for years to come.
For US AI companies, the six-week streak of Chinese dominance is a warning signal. The traditional assumption that superior benchmark performance automatically translates into market leadership is being challenged. As Chinese models continue to close the performance gap while maintaining significant price advantages, US companies will need to find new ways to differentiate their offerings beyond raw capability metrics.
As Chinese AI companies continue to refine their models and aggressively court international developers, the gap in API call volume may widen further. The data from OpenRouter shows that, in the global race for AI adoption, cost-efficiency and accessibility are proving just as important as raw technological supremacy.