GLM 4.7 Prompting Best Practices

GLM 4.7 is Z.AI's strongest open-source coding model (~358B params, ~32B active per token via MoE). This skill covers 10 rules for optimal prompting.

Deployment Options

• Cerebras cloud: For API parameters, SDK examples, and cloud setup, see references/cerebras.md
• Ollama local: For local installation, ollama launch, and VRAM requirements, see references/ollama.md

Prompt Structure

Rule 1: Front-load instructions

GLM 4.7 has a strong bias toward the beginning of the prompt (more so than other models). Place all mandatory instructions at the absolute start of your system prompt.