GLM-5.2 (High): benchmark results

GLM-5.2 evaluated at the high reasoning-effort setting. Provider: Zhipu. Released 2026-06-13. Access: Open.

Unified ELO 1681 ± 1, rank #150 of 1761 rated models, from 12 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SWE-rebench57.01Resolved (%)83.3
WeirdML67.31Average Score80.9
LisanBench0.12Mean Path Length / Current Maximum77.1
E-Commerce Bench301Final Assets (¥k, mean of 5 episodes)70.6
Opus Magnum Bench25.39Human-normalized score (%)61.9
OckBench78.5Accuracy (%)58.2
Computer Anthology Terminal Tasks (Terminus-2)31.4pass@1 (%)53.8
Surface Evolver Bench Pass Rate31.25Pass Rate (%)52
Surface Evolver Bench55.62Mean Score (%)48
CursorBench 3.151.5Score (%)23.3
Aikido CVE Rediscovery (pass@3)50Recall (%)2.2
Aikido CVE Rediscovery37.2Recall (%)0

Interactive version: theaggregate.ai/model?slug=glm-5-2-high · How It Works · Data refreshed daily, snapshot 2026-09-05.