Model Router

GateCtr picks the right LLM for each request. You pay less.

How it works

When routing is enabled, GateCtr scores each request against a set of criteria and selects the optimal model:

• Task complexity (simple Q&A vs. multi-step reasoning)

• Required output quality

• Current model pricing

• Your configured provider preferences

Simple requests go to cheaper models. Complex ones go to the best model for the job.

Enable

const response = await client.complete({
  model: 'auto', // triggers the Model Router
  messages,
});

console.log(response.gatectr.model_used); // e.g. "gpt-3.5-turbo"

Or enable globally:

const client = new GateCtr({
  apiKey: process.env.GATECTR_API_KEY,
  route: true,
});

Routing logic

Request type	Typical selection
Simple Q&A, short tasks	gpt-3.5-turbo, mistral-small
Summarization, classification	gpt-4o-mini, claude-3-haiku
Complex reasoning, code	gpt-4o, claude-3-5-sonnet

Configure provider preferences

In the dashboard: Settings → Model Router

You can restrict routing to specific providers or exclude models entirely.

Response fields

"gatectr": {
  "model_used": "gpt-3.5-turbo",
  "model_requested": "auto",
  "routing_reason": "low_complexity",
  "cost_usd": 0.00008
}

Available on

Pro plan and above.