On June 16, 2026, Z.ai (Zhipu AI) released GLM-5.2, a 753B-parameter Mixture-of-Experts model under the MIT license. It's the first open-weights model to make Claude Opus 4.8 look expensive without making it look slow.
The benchmark story is compelling:
SWE-bench Pro: GLM-5.2 scores 62.1, outperforming GPT-5.5 at 58.6
FrontierSWE: Trails Opus 4.8 by just 1%, surpassing GPT-5.5
Terminal-Bench 2.1: Scored 81.0, up from 62.0 on GLM-5.1 — a ~19-point generational leap
AIME 2026: 99.2 vs Opus 4.8's 95.7 on olympiad-grade math
The model features a 1M-token context window, max output of 131K tokens, and introduces "IndexShare" — a sparse attention mechanism that cuts per-token FLOPs by 2.9x at long context. Day 0 adaptation to domestic Chinese computing platforms including Huawei Ascend, T-Head, Moore Threads, and Cambricon means "Chinese chips running Chinese models" is now a reality.
Top comments (1)
Zhipu AI matches Claude Mythos on vulnerability detection.