BREAKING
Z.ai Releases Open Model GLM-5.2
0
B
parameters
0
B
active
0
M
context tokens
Terminal-Bench 2.1 Scores
Opus 4.8
85
GLM-5.2
81
GLM-5.1
62
0
SWE-bench Pro
0
%
FrontierSWE
Open Weights Raise Misuse Risk
Commercial API
●
Provider detects misuse
●
Can block bad actors
Self-Hosted
●
Zero provider visibility
●
Guardrails easily removed
●
Cost about half of US models
Five Eyes Warn of AI Cyber Threat
AI NEWS BLITZ
Chinese startup Z.ai has launched GLM-5.2, an open-weight model matching top US systems.