GLM-5.2: Built for Long-Horizon Tasks
Open paper
Benchmark
Model
Rank
Results
code-generation-on-swe-bench-pro
GLM-5.2
#2
Resolved: 62.1
code-generation-on-terminal-bench-2-1
Claude Code + GLM-5.2
#1
Accuracy: 82.7
Rank counts only results with a code link.