Operations

What Is Model Routing?

Directing different agent tasks to different LLM models based on complexity, cost, or speed requirements.

In Plain English

Not every task needs the most powerful (and expensive) model. A simple FAQ lookup works fine with Claude Haiku ($0.25/M tokens). A complex reasoning task needs Claude Sonnet ($3/M tokens). Model routing directs each task to the appropriate model.

This optimizes the cost-quality tradeoff. Simple tasks stay cheap. Complex tasks get the intelligence they need.

Why It Matters for OpenClaw

LLM costs can spiral without model routing. Running every task on GPT-4 or Claude Opus is expensive. Model routing cuts costs 50-80% without sacrificing quality where it matters.

How Clawctl Helps

Clawctl supports BYOK with multiple providers. Configure different models per agent in a multi-agent setup. Use fast models for triage and powerful models for complex reasoning.

Try Clawctl — 60 Second Deploy

Common Questions

Which models should I use for what?

Haiku/GPT-4o-mini for triage, categorization, and simple tasks. Sonnet/GPT-4 for complex reasoning. Opus for the hardest problems.

Can I switch models without redeploying?

Yes. Update the model in your openclaw.json or Clawctl dashboard. Takes effect immediately.

Does Clawctl support local models?

Yes. Ollama integration lets you run local models for sensitive or cost-optimized workloads.