#GLM-5.2
1 published article
Japan AISI tests Claude Opus 4.8 and open-weight GLM-5.2 on ExploitBench and finds open-model cyber spread still limited
J-AISI's 2 October notes put GLM-5.2 behind both Opus models on 41 V8 exploit tasks, at lower token cost. A standout Opus 4.8 result did not repeat on a retry.