Zhipu launched today and open-sourced GLM-5.2, specifically designed for long-term task capabilities

BlockBeats News, June 17 — Zhipu (02513.HK) announced today that it has launched and open-sourced GLM-5.2. On CodeArena, a front-end development evaluation system where millions of users worldwide participate in blind testing, GLM-5.2 delivered the top performance among globally available models. Today, GLM-5.2 is built specifically for long-range task capabilities. New features include:

Solid 1M context, reliably supporting long-range tasks; stronger sense of presence and more practical coding capabilities; extreme infrastructure optimization, running on domestic computing platforms from Day 0; MIT open-source license, no regional restrictions, and technology equality without borders.

The 1M context forms GLM-5.2’s long-range delivery capability. Multiple long-range task benchmark tables show that GLM-5.2’s performance falls between Claude Opus 4.7 and 4.8, making it the highest-ranked open-source model.

GLM-5.2’s online inference relies on multiple domestic computing platforms. It completed inference adaptation on Day 0 with platforms including Huawei Ascend, Pingtouge, Moore Threads, Cambrian, Kunlun Chip, Muxi, Hygon, and Biren. On domestic chip clusters, it achieves stable operation with high throughput, low latency, and large-scale concurrency. It is expected that after the Ascend 950 supernode is launched in the second half of the year, it will also become a strong computing foundation for GLM-5.2.

Click the original link below to join the Dongcha Beating · Feishu AI News channel and monitor global AI hotspots and news 24/7.