Chinese models sweep top four of OpenRouter global weekly token ranking; DeepSeek-V4-Flash tops at 8.83 trillion tokens
For the week of August 3-9, the top four on OpenRouter are Chinese models: DeepSeek-V4-Flash GA, Tencent Hy3, DeepSeek-V4-Flash preview, and Xiaomi MiMo-V2.5.
On August 11, reporters reviewing the OpenRouter global model API aggregator real-time weekly ranking found that Chinese-developed models took the top four spots by call volume.
DeepSeek-V4-Flash-0731 leads with 9.39 trillion tokens, Tencent Hy3 follows at 8.94 trillion, DeepSeek-V4-Flash-0423 ranks third, and Xiaomi MiMo-V2.5 fourth. Chinese models took six of the top ten, totaling around 35.58 trillion tokens, exceeding US models for 15 consecutive weeks. Data for the week of August 3-9 shows global AI model call volume at 69 trillion tokens (up 21.48% week-on-week); Chinese models at 34.25 trillion (up 21.76%), exceeding US for 15 consecutive weeks; US at 9.17 trillion (up 109.36% in a first major jump), with China still at 3.7x the US.
The top four globally are all Chinese: DeepSeek-V4-Flash GA 8.83 trillion (+570%), Tencent Hy3 8.05 trillion (+67%), DeepSeek-V4-Flash preview 5.88 trillion (-19%), Xiaomi MiMo-V2.5 5.39 trillion (-14%). On July 31, DeepSeek announced the V4-Flash GA API had entered public beta, with the new version retrained to strengthen code understanding, tool calling, and long-horizon task capability.
The commercialization inflection point also emerged the same week: on August 6 DeepSeek preannounced an across-the-board API price increase, expected to be substantial, with the market generally expecting September rollout and increases of 35%-50%. According to Reuters, Alibaba also plans to introduce revenue sharing for the next-generation Qwen3.8-Max — weights remain open, but large commercial users will need to share a portion of revenue with Alibaba.
The DeepSeek official site shows V4 Flash at 1.9 RMB per million output tokens. OpenRouter previously disclosed that agent workloads average around 15x the token consumption of ordinary human-machine interaction per request. Kaiyuan Securities interprets the DeepSeek price increase as reflecting demand growth, inference capacity, and cost recovery pressure; Alibaba's revenue share extends the commercialization boundary from API fees to commercial deployment in customer-owned data centers.