Huatai Securities said July iterations of Kimi and Q2 updates to GLM validate the core domestic-model thesis and could shorten iteration cycles. The firm estimates the technical gap versus closed‑source overseas models has narrowed to about 3–4 months; combined with higher cost‑effectiveness, domestic providers are lifting capability ceilings and pricing power. Huatai’s tests show K3 is only marginally weaker than GPT‑5.6 Sol, while GLM 5.2 benchmarks roughly at OPUS 4.7. Investment takeaway: rapid capability improvement should reinforce industry-chain positives. Near term, pricing power has improved but model‑level competition is intensifying; longer term, rising Agent penetration should expand the addressable market.(金十数据)