DeepSeek V4-Flash Tops OpenRouter's Weekly Model Rankings With 7.22 Trillion Tokens as Chinese Models Dominate Global Usage

DeepSeek's V4-Flash model took the No. 1 position in OpenRouter's weekly model-call rankings for the week of July 27–August 2, processing 7.22 trillion tokens on the aggregator platform, TechNode and EqualOcean reported August 5, following the model's July 31 '0731' pricing and performance update. Chinese models claimed four of the top five OpenRouter slots for the week — DeepSeek V4-Flash, Xiaomi MiMo-V2.5 (5.1 trillion tokens), Tencent Hunyuan Hy3 (5.01 trillion), and a second DeepSeek V4-Flash build (3.45 trillion) — with OpenAI's GPT-5.6 Luna a distant fifth at 2.99 trillion tokens. OpenRouter's rankings track only traffic routed through the aggregator, excluding direct API calls and private deployments, but are closely watched as a proxy for real-world developer adoption trends. The result extends a run of milestones for DeepSeek's cost-efficient Flash line just days after the official V4-Flash release (July 31) intensified the 'race to zero' in frontier-model API pricing.