Chinese open-source AI models have achieved top rankings in a global list, according to LMArena, a benchmarking platform created by researchers at the University of California, Berkeley.

The rankings placed China’s Kimi K2, MiniMax M1, Qwen 3, and DeepSeek R1 variant ahead of models from Google and Meta.

LMArena assessed the AI models by comparing their responses and evaluating their performance.

The highest-ranked model, Kimi K2, was launched by the Chinese startup Moonshot AI on July 11, 2025. It has received attention for its effective responses.

DeepSeek R1-0528, a refined version of a model introduced by Hangzhou-based DeepSeek earlier this year, secured second place. The platform highlighted its strengths in multi-turn dialogue and reasoning tasks.

Alibaba’s Qwen 3 series, noted for its 235 billion total parameters, ranked third for its raw reasoning power.