National Model Throne Changes Hands Again? DeepSeek Returns to the Top with a Single Flash

Beating AI News Flash: Over the past few months, the throne of China's top domestic model has been alternately occupied by Kimi and Zhipu. In the latest third-party leaderboard from Artificial Analysis, Kimi K3 and GLM-5.3 scored 44 and 45 points respectively, while DeepSeek V4 Pro 0813 scored only 36 points.

Now DeepSeek has struck back with V4.1 Flash. According to officially published benchmarks, among the 16 tests where both it and Kimi K3 have results, V4.1 Flash won 13; among the 13 tests where both it and GLM-5.3 have results, it won 11. Its advantages are mainly concentrated in Coding, terminal operations, and Agent automation.

And this may only be an appetizer — after all, Kimi K3 is Kimi's current strongest 2.8T model, and GLM-5.3 is also Zhipu's flagship; yet V4.1 Flash is merely the smallest model in DeepSeek's new architecture series.