Chinese President Xi Jinping used the World AI Conference in Shanghai on Friday to cast Beijing as the natural convener of an alternative global AI order, hours after Chinese startup Moonshot AI released Kimi K3, a 2.8-trillion-parameter model the company says beats Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5 on some benchmarks.
The same-day one-two delivered a coordinated pitch: China intends to compete with U.S. labs on frontier models and to write the rules that govern them. Xi's speech followed Thursday's signing by 29 countries of an agreement to create the World Artificial Intelligence Cooperation Organization, or WAICO, a Shanghai-headquartered bloc that Beijing has cultivated since Premier Li Qiang unveiled the concept a year ago.
The model
Moonshot said Kimi K3 is China's largest AI model to date and consistently outperformed other tested models, though it still trails Anthropic's Claude Fable 5 and OpenAI's GPT 5.6 Sol on overall performance. Beijing-based Moonshot, founded in 2023 and backed by Alibaba and Tencent, raised $2 billion in May at a valuation above $20 billion, Bloomberg reported.
"Despite persistent hardware/compute capacity constraints in China, K3 demonstrates that pre-training scaling, paired with architectural innovation, can still deliver step-change gains for flagship Chinese models," Bank of America analysts led by Alex Liu wrote in a note.
Chinese AI rivals sold off. Z.ai, which launched a model to fanfare in June, dropped 28 percent on Friday. MiniMax Group fell 16 percent. Alibaba, which markets its Qwen series as the open-source leader among Chinese labs, slipped 4 percent, giving back gains from its Apple partnership earlier in the week.
The summit
Xi announced that China will offer developing countries 5,000 opportunities in AI training and seminar programs, and will pursue AI cooperation with the Association of Southeast Asian Nations, the League of Arab States and the African Union. Al Jazeera reported that WAICO's founding members include Indonesia, Brazil, Malaysia, South Africa, Senegal, Russia and Pakistan.
"AI development should not be a solo performance by a single country, but a symphony of international cooperation," Xi said, according to a Google translation of his Mandarin remarks. He urged countries to oppose "overstretching the national security concept in the field of AI or placing one country's security over that of others" — a line widely read as directed at Washington, though Xi did not name the United States.
Chip walls
The pitch coincided with fresh evidence that U.S. export controls have carved a hole in China's supply of advanced chips. Nvidia said in its annual report that it was "effectively foreclosed from competing in China's data center computing/compute market" as of the end of fiscal year 2026, warning that its exit had let Chinese rivals "build larger developer and customer ecosystems to challenge us worldwide." At the Shanghai event, Huawei showcased its Atlas 950 SuperPoD, a chip-linking supernode designed for large-model training.
Moonshot's benchmark claims have not been independently verified, and the comparisons the company chose — Claude Opus 4.8 and GPT 5.5 — sit, by its own accounting, one rung below the U.S. leaders it did not test against. Xi's pledge that AI must remain "always under human control" also sidesteps Beijing's own record of censoring model outputs and requiring approval before public deployment. Governance expert Arindrajit Basu, writing for the Carnegie Endowment for International Peace, said Beijing is using the alliance to seek "buy-in for its state-centric governance vision of technology from the Global South."
Xi is scheduled to meet President Trump in September, after a year of trade-war escalation. Alex Liu's note put the market reading plainly: "K3 raises the capability ceiling for China AI models, shifting the burden of proof to other independent AI labs."

