Zhipu Launches and Open Sources GLM-5.2, Designed for Long-Range Task Capabilities

On June 17, Zhipu (02513.HK) announced the launch and open-sourcing of GLM-5.2. In the CodeArena front-end development evaluation system, which involved blind testing by millions of users globally, GLM-5.2 achieved the top performance among globally available models. Designed specifically for long-range task capabilities, GLM-5.2 features include: Solid1M context for stable support of long-range tasks; enhanced usability and practical coding abilities; extreme infrastructure optimization, running on domestic computing platforms from Day 0; and an MIT open-source license with no geographical restrictions, promoting technological equality across borders. The 1M context endows GLM-5.2 with long-range delivery capabilities, with multiple long-range task benchmarks indicating that its performance ranks between Claude Opus 4.7 and 4.8, making it the highest-ranked open-source model. The online inference of GLM-5.2 relies on several domestic computing platforms, having completed inference adaptation with Huawei Ascend, PingTouGe, Moore Threads, Cambricon, Kunlun Core, Muxi, Haiguang, and Birun on Day 0, achieving stable operation with high throughput, low latency, and large concurrency on domestic chip clusters. It is expected that after the launch of the Ascend 950 super node in the second half of the year, it will also serve as a strong computing foundation for GLM-5.2. --------------------------------- Click the link below to join the Beating · Feishu AI news channel, which monitors global AI hotspots and news 24/7.