AsianFin --Three Chinese-developed AI models have taken the top spots in the global open-source large model rankings, with Kimi K2 securing first place, according to LMArena, a leading platform known for its dynamic blind testing by thousands of developers.

Kimi K2, developed by Chinese AI startup Moonshot AI, edged out major Western models including Google’s Gemma 3 and Meta’s Llama 4. DeepSeek R1 and Qwen3 — developed by DeepSeek and Alibaba’s Qwen team respectively — ranked second and third, marking a clean sweep of the top three positions by Chinese models.