Moonshot and Alibaba release open-source models rivaling GPT-5.6 and Claude Fable 5

In a major escalation of the AI arms race, two of China's leading AI companies have released frontier models that they claim can go head-to-head with the best from OpenAI and Anthropic at significantly lower cost. Beijing-based Moonshot AI unveiled Kimi K3 on Friday, a 2.8 trillion parameter open-source model that it says consistently outperforms nearly all US systems except OpenAI's GPT-5.6 Sol and Anthropic's Claude Fable 5. Over the weekend, Alibaba previewed Qwen3.8, a 2.4 trillion parameter model it describes as 'second only to Fable 5.' Both companies are emphasizing a key difference from US labs: their most advanced models are publicly available for download and modification.

The rapid-fire releases, coming just months after DeepSeek shocked the industry with a low-cost rival, suggest America's lead at the AI frontier is narrowing. The Chinese models are open-weight or open-source, in stark contrast to the proprietary approach of OpenAI and Anthropic, whose most capable systems remain locked behind APIs. This openness is particularly notable as Washington imposes export controls to restrict China's access to advanced chips and even forced Anthropic to pull a capable system from the market. Whether Kimi K3 and Qwen3.8 fully live up to their claims remains to be seen—independent testing is pending—but the releases are already sharpening the geopolitical and economic stakes in the defining technological race of our time.