CAISI's latest report, released this week, puts Moonshot AI's Kimi K3 firmly in fourth place on major intelligence indices, behind the current American frontrunners.
K3 is no small project. It runs on 2.8 trillion parameters, carries an open-weights architecture, and supports a 1-million-token context window, which puts it among the most ambitious models China has shipped to date.
Still, Moonshot's own technical assessments acknowledge the gap. K3 trails both Claude Fable 5 and GPT-5.6 Sol on overall performance, and CAISI's independent analysis lines up with that conclusion.
The broader picture hasn't shifted much. US models continue to hold the top spots in frontier AI, even as Chinese labs have visibly closed the distance over the past year.
One market signal worth noting: prediction-market pricing has drifted toward lower odds for Alibaba leading the AI sector by the end of August 2026, suggesting traders aren't yet convinced China will overtake the US within that window.
The race is tightening, but the finish line keeps moving. New benchmark releases or model updates from US companies before mid-2026 could easily reset the competitive picture again.
