According to a research report, the continuous iteration of domestic large language models will effectively expand the application scenarios for domestic AI and accelerate the formation of a commercial ecosystem within China's AI landscape. The expansion of AI commercial applications is central to the technology sector's growth trajectory, and the deployment of AI agents with long-term task capabilities is expected to drive new commercial scenarios. The accelerated iteration of domestic large models will fuel rapid growth in demand for domestic computing power and switching chips. The main points from the report are as follows: Kimi K3 is the world's first open-source model at the 3 trillion parameter scale, demonstrating outstanding performance across various task scenarios.
Programming: K3 possesses robust long-context coding abilities, enabling it to complete extended, complex tasks with minimal human supervision, showcasing excellent long-term task execution.
Knowledge Work: Kimi K3 shows consistent advantages across different production-oriented workflows, with a significant enhancement in its agent-based knowledge work capabilities. For example, in industry research, K3 underwent over 120 rounds of recursive self-improvement to generate a report on inference chips, conducting more than 2,800 web searches and scrapes, over 1,100 terminal data pulls, and processing 87 quarterly reports and 99 original PDFs, totaling more than 11,000 pages of material.
The report posits that the ongoing performance iteration of domestic large models will effectively broaden domestic AI application scenarios, speeding up the commercial cycle within China's AI ecosystem.
Long-Term Tasks and Self-Evolution Continuously Expand AI Commercial Applications
According to Tang Jie, founder of Zhipu AI, long-term task capabilities, fully autonomous intelligent agent systems, and self-evolution are core future directions for AI. Future AI will possess planning and execution capabilities spanning weeks, months, or even years, and autonomous, collaborative, 24/7 operational agent collectives will become new forms of productivity. The report believes the expansion of AI commercial scenarios is the core driver for the AI technology sector, and the implementation of AI agents with long-term task capabilities is anticipated to catalyze the emergence of new commercial applications.
Rapid Iteration of Domestic AI Models Drives High Growth in Domestic Chip Demand
The accelerated iteration of domestic AI large models, with Kimi K3's overall performance stably surpassing all models except Claude Fable 5 and GPT-5.6 Sol, and the imminent official release of the 2.4 trillion parameter Qwen 3.8, leads the report to conclude that this iteration will drive high growth in demand for domestic computing power and switching chips.
Risk Factors
Potential risks include a slower-than-expected pace of technological iteration for domestic AI large models; slower-than-anticipated progress in AI commercialization; inadequate supply of raw materials; risks of product price reductions; and slower-than-expected improvements in chip manufacturing yield rates.