LLM battle heats up in July: domestic performance rises rapidly as overseas giants launch a price war

At present, the competitive landscape for global frontier models is marked by a tense situation of “alternating leadership, with the endgame still undecided.” Google briefly topped the charts at the end of 2025 thanks to Gemini 3’s native multimodal capabilities. In early 2026, Anthropic gained the upper hand by leveraging outstanding Coding (programming) abilities and an agent product flywheel driven to the extreme, while OpenAI reclaimed the throne of frontier intelligence in the second quarter of 2026 through a limited release of GPT-5.5/5.6.

Amid this structural window period, in which the world’s top closed-source vendors are constrained by compute supply bottlenecks, buffeted by recurring “system de-intelligence” incidents, and burdened by high pricing strategies, domestic AI large models are carving out the market in an exceptionally powerful and pragmatic stance, kicking off a sweeping historical process of a “feudal lords’ war” and “domestic models going global.” China’s domestic open-source and closed-source ecosystems are rapidly transforming from “followers” on the global AI map into “rule rewriters,” driven by algorithmic innovation (such as DeepSeek’s extreme cost reconstruction), deep cultivation of vertical scenarios (such as the Coding ecosystem behind the Zhipu GLM series), and consumer-grade multimodal breakthroughs (such as the video-giant pattern involving ByteDance, MiniMax, and Keling).