Moonshot AI's Kimi K3 narrows China-US AI gap, shaking Wall Street

Moonshot AI, a rising Chinese tech company, has unveiled Kimi K3, a 2.8-trillion-parameter AI model that has triggered intense discussion across global markets and forced Wall Street to rethink the pace at which Chinese labs are closing the capability gap with the West. The model scored 57 on the independent Artificial Analysis Intelligence Index, placing it within striking distance of America's leading models: Claude Fable 5 and GPT-5.6 Sol.

Morgan Stanley analyst Gary Yu views K3 not as an isolated breakthrough but as evidence of cumulative progress across China's AI industry. "K3 has received positive feedback globally, signaling an all-round catch-up of Chinese LLMs with US leaders in model size, performance, and pricing," Yu said. He noted that K3 is the largest open-weight Chinese LLM so far and suggested the release may indicate that the scaling law still holds. Yu explicitly rejected the overnight-miracle narrative: "We do not view K3 as an overnight miracle but rather as the result of cumulative progress across China's AI model industry." He pointed to GLM-5.2, released in June, as evidence of this steady momentum and predicted that more Chinese LLMs with larger sizes, higher pricing, and better performance would follow with strong global competitiveness.

Bernstein analyst Robin Zhu characterized K3 as "a home run" and called attention to a pattern of Chinese innovation that has become impossible to ignore. After DeepSeek V3 and GLM-5.2 earlier in the year, K3 represented another instance where China's top AI labs surprised global investors by keeping pace with the US frontier. Zhu highlighted that K3 demonstrated impressive reasoning and asset generation capabilities, with emphasis on the aesthetics of deliverables. He noted that immediate market reactions to K3—including declines in China AI lab stocks and semiconductor stocks—felt sensible given the structural shift in competitive dynamics.

For several quarters, Wall Street consensus held that US AI firms enjoyed a comfortable, multi-month lead in frontier capabilities. Kimi K3 represents a turning point that has forced analysts to substantially revise this view. Morgan Stanley's Gary Yu frames the release not as an overnight breakthrough but as evidence of steady momentum—the latest in a series of strong Chinese model releases, including GLM-5.2 in June. This pattern suggests that the scaling law (the principle that larger models perform better) continues to hold and that Chinese labs have cracked the core technical and resource challenges needed to compete at the frontier.

Bernstein's Robin Zhu went further, explicitly noting that after DeepSeek V3 and GLM-5.2, K3 now represents a third instance of Chinese labs surprising global investors by keeping pace with US frontier models. The analyst's emphasis on reasoning and asset generation capabilities—dimensions where K3 showed particular strength—indicates that Chinese competitiveness is not confined to raw benchmark scores but extends to practical model quality. Market reactions cited by Zhu (declines in China AI lab stocks and semiconductor stocks) appear to reflect investor recognition that the competitive landscape has structurally shifted.