Overview
- The Beijing summit on Jan. 10–11, organized by Tsinghua Beijing Key Lab and Zhipu AI, gathered leading Chinese AI figures to outline current directions for models and agents.
- Tencent chief scientist Yao Shunyu, in his first public remarks since joining the company, said consumer use resembles enhanced search while enterprise deployments are transforming workflows and favoring the strongest models.
- Yao pointed to ChatGPT as a consumer exemplar and Claude Code for enterprise use, adding that higher intelligence in ToB settings maps directly to productivity and revenue.
- Zhipu founder Tang Jie said DeepSeek’s early‑2025 release catalyzed a shift toward AI that does tasks and reported that GLM‑4.5 integrated coding, reasoning and agents, with GLM‑4.7 delivering large Agent and Coding gains over 4.6 and 4.5, according to company benchmarks.
- Qwen technical lead Lin Junyang said the model itself must be treated as a product and described agents progressing toward long-duration, environment-interactive and potentially physical-world work.