Moonshot AI Debuts 2.8 Trillion-Parameter Kimi K3 as Chinese Rivals’ Shares Slide
Updated
Updated · CNBC · Jul 17
Moonshot AI Debuts 2.8 Trillion-Parameter Kimi K3 as Chinese Rivals’ Shares Slide
3 articles · Updated · CNBC · Jul 17
Summary
Moonshot AI said Friday its new Kimi K3 model outperformed OpenAI’s GPT 5.5 and Anthropic’s Claude Opus 4.8 on coding and agent benchmarks, though it still trails GPT 5.6 Sol and Claude Fable 5 overall.
The 2.8 trillion-parameter model is China’s largest so far, and Bank of America said it shows Chinese labs can still post major gains despite domestic hardware and compute constraints.
Chinese AI stocks fell after the release, with Z.ai plunging 28%, MiniMax dropping 16% and Alibaba slipping 4% as analysts said K3 raises the bar for independent model builders.
The launch adds pressure in the U.S.-China AI race as Chinese models gain Western users by narrowing performance gaps while staying cheaper than top American systems, drawing scrutiny from U.S. lawmakers.