Overview
- The AGI-Next summit, organized by Tsinghua University’s Beijing Key Laboratory with Zhipu AI, brought together industry and academic leaders to outline China’s next phase of AI development.
- In his first public remarks since joining Tencent, chief scientist Yao Shunyu said consumer uses still feel like enhanced search, whereas enterprise adoption is accelerating as teams code and operate systems through natural language.
- Yao argued that in business settings higher model intelligence directly boosts productivity and revenue, pushing companies to choose the strongest models available.
- Zhipu AI founder Tang Jie framed DeepSeek’s 2025 debut as a shock that shifted focus from chat to getting work done, and said GLM-4.5 integrated coding, reasoning, and agent capabilities with leadership on a dozen domestic benchmarks.
- Tang reported that the recent GLM-4.7 release delivered large gains in Agent and Coding performance over GLM-4.6 and 4.5, while Qwen’s Lin Junyang stressed that foundation models must be built as real products and described agents evolving toward long-duration and eventually embodied tasks.