The research report from CITIC Construction Investment highlights the rapid iteration and enhancement of local models, exemplified by advancements in GLM-5.2, Kimi K2.7 Code, and others. These models have seen significant improvements in capabilities such as handling 1M context lengths, long-range agent functionality, agentic coding, and real-world engineering deployment. This progress is propelling the transformation of local models from general-purpose question-and-answer tools to sophisticated developer instruments and enterprise-level workflow solutions. Kimi, in particular, has bolstered its international operational prowess, while DeepSeek's recent financing round has heightened expectations for the industrialization of cutting-edge models.

The gray-scale testing of WeChat AI suggests a strategic shift in AI entry points, moving away from standalone applications towards integration within super-app ecosystems. This transition is anticipated to boost the frequency of model invocations and increase the consumption of reasoning tokens. On the global stage, Anthropic is experiencing sustained growth in revenue, large customer acquisition, and financing scale, driven by robust demand for enterprise workflow solutions and agentic coding services.

In the realm of computing power, AI cloud providers are witnessing a surge in order backlogs, coupled with rising GPU rental prices. Supply constraints for CPUs, DRAM, and NAND are also intensifying, indicating that the markets for computing power, storage, and leasing services are poised for continued prosperity.