Moonshot AI, a Chinese tech firm, has unveiled Kimi K3, its first open AI model surpassing 2.8 trillion parameters.The model features a 1-million-token context window and claims a 2.5× improvement in scaling efficiency over its predecessor, Kimi K2.While acknowledging that Kimi K3 lags behind U.S.proprietary models like Claude Fable 5 and GPT 5.6 Sol, Moonshot highlights its strengths in coding, knowledge work, and multimedia creation.
The announcement comes amid growing concerns about Chinese AI firms leveraging cheaper rates and data distillation techniques to outcompete Western counterparts.Anthropic has accused companies like DeepSeek and MiniMax of illicitly extracting Claude’s capabilities.Moonshot emphasizes its commitment to open-source collaboration while balancing proprietary advancements, backed by tech giants Alibaba and Tencent.The release of full model weights by July 27 underscores China’s push to lead in AI innovation, aligning with President Xi’s 2035 sci-tech goals.
Original title: What is Kimi K3, China’s first open AI model to reach 2.8 trillion parameters?
The AI system has determined that this news is clickbait/sensationalist: : The original title uses hyperbolic language like 'first open AI model' and 'reach 2.8 trillion parameters' to attract attention, which is more sensational than factual. This has coincided with the opinion of the majority of users.