Chinese AI Models Dominate OpenRouter Usage

In July, Chinese-developed AI models achieved a significant milestone by occupying all five top positions on OpenRouter, a neutral routing platform that serves as a key indicator of AI industry adoption. Xiaomi's MiMo V2.5 led in token volume, followed by models from DeepSeek, MiniMax, Alibaba's Qwen family, and Moonshot's Kimi. These Chinese models now account for over 60% of the platform's weekly traffic, which exceeds 20 trillion tokens. This represents a substantial shift from a year prior, when US models dominated approximately 70% of OpenRouter's traffic, compared to their current 30%.

The trend is further underscored by the fact that by mid-July, Chinese models processed a record 58% of tokens used by American firms on the platform. This adoption is driven by economic factors, as US companies are increasingly choosing Chinese AI solutions based on their compelling price-performance metrics. While American AI labs continue to lead in absolute frontier capabilities, with models like GPT 5.5, Claude Fable 5, and Gemini 3.x excelling in complex reasoning, long-horizon agents, and demanding enterprise tasks, the AI race has bifurcated into capability and distribution. The United States is currently leading in capability, but losing ground in distribution.

Cost efficiency is a major differentiator. DeepSeek's V4-Pro model is priced at approximately one-twelfth the cost of GPT-5.5 for comparable benchmark performance. Specifically, DeepSeek V4 Flash costs $0.14 per million input tokens, a stark contrast to GPT-5.5's $5.00 per million input tokens. OpenRouter's internal analysis indicates that Chinese open-source models are 60% to 90% cheaper than leading American offerings. For high-volume production workloads such as coding agents, document processing, and customer operations, this cost differential is a decisive factor in purchasing decisions.

Distribution is crucial for establishing ecosystem lock-in. Alibaba's Qwen family of models has surpassed one billion cumulative downloads, overtaking Meta's Llama as the most downloaded open-source model family globally. Llama had previously set the standard for open-weight AI models. The shift in usage on OpenRouter, particularly the increasing reliance of US firms on Chinese AI for production workloads, highlights a strategic challenge for American AI companies. The economic advantages offered by Chinese models are proving irresistible for businesses focused on cost optimization and scalability in their AI deployments. This trend suggests a potential rebalancing of global AI market share, driven by practical economic considerations rather than solely by frontier technological advancements.