[DigitalToday reporter Chi-gyu Hwang] Chinese AI models have far outpaced U.S. models in token usage, India’s business outlet NDTV Profit reported on the 29th (local time), citing OpenRouter data.

OpenRouter, a routing platform that integrates AI model calls (APIs) to allow comparison and use, said Chinese AI models have overtaken U.S. models in token consumption so far this year. NDTV Profit reported that efficient model design allowed Chinese AI research companies to set token prices lower than leading U.S. firms, and that this has emerged as a new advantage in the AI race.

Among the top 5 models by token usage, only one U.S. company, Anthropic, was represented. The most-used model was DeepSeek V4 Flash with 4.63 trillion tokens, followed by MiniMax M3 with 4.13 trillion tokens and Xiaomi MiMo-V2.5 with 3.8 trillion tokens.

Tencent Hy3, Anthropic Claude Opus 4.7, DeepSeek V4 Pro, Owl Alpha, Claude Sonnet 4.6 and Claude Opus 4 were included in the top 10. Google Gemini 3 and OpenAI GPT-5.5 placed 12th and 13th, respectively.

OpenRouter said weekly token usage for the week ended June 15 rose to 47.1 trillion tokens from 44.6 trillion tokens the previous week. On an individual model basis, DeepSeek accounted for the largest share.

The gap between U.S. and Chinese models emerged as several U.S. AI companies switched pricing plans from flat-rate subscriptions to token-based billing. Usage-based pricing led corporate customers to feel AI tool costs more directly, and Amazon, Cisco, Uber, Walmart and Meta began controlling costs by setting caps on AI usage or steering users toward cheaper models.