In the past week, the total invocation volume of AI large models worldwide surged to 28.9 trillion Tokens, marking the fifth consecutive week of growth. Notably, the weekly invocation volume of Chinese large models reached 9.223 trillion Tokens, a significant 19.89% increase from the previous week. This performance has allowed Chinese AI to outpace the US for four consecutive weeks, securing its position as the global leader.

In terms of model rankings, DeepSeek-V4-Flash emerged as the frontrunner, closely followed by Tencent's Hy3 preview. DeepSeek boasted three models among the top nine, with a combined invocation volume of 5.74 trillion Tokens, maintaining its lead for two consecutive weeks. On the other hand, Kimi K2.6, which previously held the sixth position, failed to sustain its ranking and dropped out of the list.

A notable development is DeepSeek's announcement of API discounts for its V4-Pro model, with the price set to be reduced to one-fourth of the original. This pricing strategy not only spurred a substantial increase in its invocation volume but also prompted industry-wide discussions on price competition and profitability models.