AUGUST 5TH MESSAGE, THE LARGEST AI LARGE MODEL IN THE WORLD Aggregation Platform OpenRouter LAST WEEK, D@eepSeek V4 Flash is at the top of the platform with a 7.22 trillion token weekly call, with a statistical cycle from 27 July to 2 August, an increase of 13% over the previous week。
According to OpenRouter data, the two versions of DeepSeek V4 Flash processed a total of 67 billion tokens, with a total of 329 trillion tokens per week and a total of 61.3 trillion tokens per week。
Among the top 10, the Chinese models DeepSeek, MiMo-V2.5, Hy3 and GLM-5.2 occupied many seats and most of them were free or near free. OpenAI ranked 8th in GPT-5.6 Luna, the only OpenAI model in the top 10, with an increase of 465%, but still less than one third of DeepSeek V4 Flash。
OpenCode disclosed separately that DeepSeek V4 Flash's single-day processing capacity on its platform had reached 8 trillion token, of which 5 trillion came from free trial-level consumption and 3 trillion was used by developers on a fee basis。