BEIJING, Aug. 2026 - From January 2025 when DeepSeek-R1 shocked the global AI community with world-class reasoning at a fraction of US competitors' cost, to August 2026, the Chinese AI field is experiencing an explosion of new products, rapidly closing the gap with Silicon Valley. This transformation is being called the evolution from "DeepSeek Moment" to "DeepSeek System".
From late July to early August 2026, Chinese large model companies launched new models in succession: Zhipu released GLM-5.2, Moonshot AI launched Kimi K3 with 2.8 trillion parameters, ByteDance, DeepSeek, and MiniMax all unveiled updates, and Alibaba released Qwen3.8-Max. Five major model releases occurred within five weeks, showcasing the system's competitive strength.
The 2026 "China AGI Innovation Influence Institution TOP 30" report pointed out that over the past year, the AI industry focus has returned to model innovation: DeepSeek upgraded from R1 to V4, with significantly narrowed gaps between open-source and closed-source models; Zhipu and MiniMax listed on the Hong Kong Stock Exchange, becoming the world's first batch of listed large model companies; Kimi released the 2.8 trillion parameter K3; Qwen continued multi-size multi-modal open-source.
The industry transformation has extended beyond the model layer: Coding Agents have achieved large-scale application deployment, and vibe coding is reshaping software development paradigms.
Source: BAAI Hub, Baijiahao
