A three-tier task routing logic designed to minimize API costs by escalating from Flash to Plus models only when complexity requires it.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install smart-model-routing-for-zai
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install smart-model-routing-for-zai using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Smart Model Switching (GLM)?
Smart Model Switching is a strategic routing framework designed for the z.ai (GLM) ecosystem. It implements a cost-first approach to task execution by prioritizing the cheapest available compute resource for every request. This logic is essential for developers using Openclaw Skills who want to maintain high reasoning quality without overspending on simple operations like status checks, greetings, or basic factual lookups.
By categorizing tasks into Flash, Standard, and Plus tiers, the skill ensures that expensive high-reasoning models like GLM-4-Plus are reserved for architecture and deep debugging, while lighter models handle the bulk of routine operations. This approach maximizes efficiency and performance across complex multi-agent workflows.
Smart Model Switching (GLM) Use Cases
- Automating routine factual Q&A and system heartbeats using the ultra-efficient Flash tier.
- Routing standard coding tasks, documentation writing, and multi-step analysis to the Standard workhorse models.
- Handling complex architectural decisions and multi-file refactoring by escalating to Plus or 32B tiers.
- Reducing overhead for cron jobs and monitoring scripts by defaulting to low-cost models.
How Smart Model Switching (GLM) Works
- The system evaluates the incoming task against a set of complexity rules and the Golden Rule of 30-second human thinking.
- Tasks involving simple lookups, greetings, or short status checks are assigned to the Flash tier (e.g., GLM-4.5-Flash).
- If the task requires code generation over 10 lines or multi-step reasoning, it is routed to the Standard tier.
- For tasks involving strategic planning, security reviews, or database schema design, the system escalates to the Plus/32B tier.
- Subagents are spawned with specific model parameters to ensure the right tool is used for the right job.
Smart Model Switching (GLM) Setup
To implement this model switching logic within your Openclaw Skills setup, define your subagent spawning calls with the appropriate model strings based on task type.
# Example of spawning a task using the Standard tier
sessions_spawn(task="Build the REST API endpoint", model="GLM-4.7")
For automated tasks, configure your cron payloads to specify the Flash model by default:
# Standard cron configuration for low-cost monitoring
{
"payload": {
"kind": "agentTurn",
"model": "GLM-4.5-Flash"
}
}
Smart Model Switching (GLM) Data Schema & Taxonomy
The skill utilizes a hierarchical model taxonomy to organize task distribution. This structure is typically reflected in session metadata and subagent configuration.
| Tier | Purpose | Target Models |
|---|---|---|
| Flash | Status checks, lookups, heartbeats | GLM-4.5-Flash, GLM-4.7-Flash |
| Standard | Code generation, planning, synthesis | GLM-4.6, GLM-4.7 |
| Plus / 32B | Architecture, deep reasoning, strategy | GLM-4-Plus, GLM-4-32B-128K |
Smart Model Switching (GLM) Advanced Features
- Intelligent escalation triggers: Automatically moves tasks up the tier chain if the current model struggles with complexity.
- Optimized subagent spawning: Specific session routing for modular project tasks.
- Multi-tier cron support: Ensures periodic tasks run on the most economical tier unless specific reasoning is required.
- Complexity classification: Pre-response analysis to determine if a human would need more than 30 seconds of thinking to solve the problem.
SKILL.md
Loading
Related Openclaw Skills
Crypto Scam Detector
A high-performance, database-first security tool for Openclaw Skills that provides instant detection of cryptocurrency scams and malicious blockchain addresses.
Crypto Genie
An AI-driven cryptocurrency safety assistant that provides instant, database-backed protection against blockchain scams and phishing.
Crypto Scam Detector
A high-performance, database-first security tool for identifying cryptocurrency scams and malicious addresses instantly.
x402 Payment Protocol
A protocol for instant, gasless stablecoin payments over HTTP using the 402 Payment Required status code.
Mini Diary
An AI-driven minimal diary for daily journaling and task tracking with intelligent auto-tagging and Markdown support.
Cloudsways SmartSearch
An advanced web search and content extraction skill that provides AI agents with real-time snippets, dynamic fragments, and full-text data for deep research.
Featured*
