Analysis: Domestic models accelerate iteration, computing power prosperity continues

CoinFeed, June 21 - According to a research report by CITIC Securities, domestic models continue to iterate. GLM-5.2 and Kimi K2.7 Code enhance 1M context, long-range Agent, Agentic Coding, and real engineering delivery capabilities, driving domestic models from general Q&A to developer tools and enterprise-level workflows. Kimi strengthens international operation capabilities, DeepSeek financing reinforces industrialization expectations for leading models, and WeChat AI gray-scale testing shows that AI entry points are moving from standalone apps to super-app ecosystems, which is expected to increase model call frequency and inference token consumption. Overseas, Anthropic's revenue, number of large customers, and financing scale continue to be revised upward, and enterprise workflow and Agentic Coding demand remain in an expansion channel.