According to Baidu, on May 13 during the Create 2026 developer conference, the company announced that its Kunlun Chip Tian Chi 256-card supernode will officially launch in June, with throughput performance improved by 25% compared to the previous generation and inference efficiency enhanced by 50%.
The supernode has completed adaptation for mainstream models including Wenxin, DeepSeek, GLM, and MiniMax, with end-to-end latency optimized by 50% through upgraded HPN 5.0 network architecture. It supports building clusters ranging from tens of thousands to millions of cards on demand.
Related News
Hon Hai CPO exchange cabinet delivery rumored to be advanced for Nvidia; full-optical technology with high gross margin becomes the second growth engine
Loongson ships 1M 3A6000 CPUs in China domestic chip push
Wu Jihan: GPU deployment target is 10k units; Bitdeer’s April AI Cloud annualized revenue rose 60%