GLM-5 Turbo
Proprietary
About
Purpose-built variant of GLM-5 optimized for agent orchestration and complex automated workflows. Features native agent-friendly training emphasizing tool use, command following, and persistent task execution. Designed for OpenClaw (Lobster Agent) workflows with ~0.67% tool-call error rate. Achieves 20% higher inference performance vs base GLM-5.
Capabilities
VisionMultimodalReasoningFunction CallingTool UseJSON ModeCode Execution
