BREAKING
GLM-5.2 Hits No. 10 on Agent Arena
0
#
Overall rank
0
Sessions
0
Models ranked
0
%
Confirmed Success
0
%
Praise vs Complaint
0
%
Tool Hallucination
Agentic Coding Benchmarks Jump
Terminal-Bench 2.1
81
GLM-5.1 prior
62
SWE-bench Pro
62.1
Open Weights, Wide Access
GLM-5.2 (Max)
Open
●
MIT License on Hugging Face
●
1M-token context window
●
128K-token max output
Closed leaders
Top
●
GPT-5.5 series
●
Claude Opus / Fable series
●
Near Claude Opus 4.8
Open-Closed Gap Narrows
AI NEWS BLITZ
Z.ai's GLM-5.2 has cracked the top ten on the Agent Arena leaderboard.