Kimi’s Founder of Moonshot AI Publishes Article in People’s Daily: “Let China’s Large Models Become the Foundation for Global Innovation”
TechFlow Selected TechFlow Selected
Kimi’s Founder of Moonshot AI Publishes Article in People’s Daily: “Let China’s Large Models Become the Foundation for Global Innovation”
For thousands of years, humanity has continually sought new ways to push the boundaries of knowledge. Currently, artificial intelligence shows promise as one of the “powerful levers” for such breakthroughs.
Navigating Web3 tides with focused insightsAuthor: Yang Zhilin, Founder and CEO of Moonshot AI’s Kimi
Recently, “agents” and “tokens” have become frequent topics of conversation. Seemingly niche open-source AI projects are giving rise to entirely new business models and market opportunities. In this wave of technological advancement, domestically developed large language models (LLMs) in China have made remarkable progress and expanded rapidly into diverse applications—not only injecting vitality into the intelligent economy but also gradually becoming a foundational platform for global innovation.
As a key driver of the latest scientific revolution and industrial transformation, AI is currently at a critical inflection point in its development. For instance, in programming capabilities, AI demonstrates tremendous application potential across code design, writing, testing, and deployment—reaching human-expert levels in certain knowledge-intensive tasks. The industry widely expects AI to further boost societal productivity and become an indispensable assistant in people’s work and daily lives.
In the past, discussions about China’s internet and digital economy often centered on its massive user base and diverse application scenarios—advantages that continue into the AI era. In March this year, China’s average daily token consumption surpassed 140 trillion, representing over a 1,000-fold increase in just two years; generative AI users exceeded 600 million, with a penetration rate of 42.8%. More importantly, China has cultivated a cohort of homegrown LLMs rivaling the world’s most advanced models—emerging as an undeniable variable in global AI innovation. From topping international LLM leaderboards in specific benchmarks, to advancing multimodal model capabilities, to evolving from single agents into agent clusters—China’s LLM growth manifests not merely in quantity, but more significantly in quality, depth of application, and global influence.
Last year, Chinese open-source models ranked first globally in total downloads. In our recently released technical report, we conducted the first-ever fundamental reconstruction—in ten years—of the residual connection, a core component of the Transformer architecture, opening up entirely new evolutionary pathways for model training and sparking widespread discussion both domestically and internationally. This breakthrough not only enables broader classes of domestic chips to be practically deployed for LLM inference, but also reduces overall LLM operational costs.
Numerous Chinese open-source models combine quantitative scale with qualitative excellence, increasingly serving as “benchmarks” in global model evaluations and vital reference samples for overseas research institutions—thereby expanding their global influence. Our team’s domestically developed LLM, Kimi, distinguished by its high performance-to-cost ratio, has secured a foothold in global markets and become a “productivity tool” for many overseas developers and users, with clear growth in paid API calls and other usage metrics. We believe, unlike closed-source approaches, China’s open-source models can achieve broad adoption and foster a self-sustaining ecosystem—propelling global progress through collective effort (“many hands make light work”).
Rapidly advancing LLMs require time to mature and refine—demanding sustained dedication and deep technical perseverance, epitomized by the Chinese saying “willingness to sit on a cold bench for ten years.” Throughout our technical攻坚 (tackling tough challenges), we continually encounter problems—and each time one is solved, another emerges. Only by calmly confronting problems head-on and diligently resolving them can we leave a series of solid, forward-moving footprints along AI’s long journey of exploration—and ultimately reap abundant rewards after becoming “friends of time.”
For thousands of years, humanity has persistently sought new methods to push the boundaries of knowledge. At present, AI shows strong promise as one of the most powerful levers for such breakthroughs. Through rigorous foundational technological innovation and an open mindset, we aim to deliver inclusive AI technologies to the world—and contribute more Chinese strength to global innovation.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
