Smart Router (model-router)

model-router is an auto-router: it inspects each request and forwards it to the best-fit underlying model (GPT-5.x, Claude, Grok, DeepSeek, …), then returns the model it actually used in the response's model field.

Usage

Just send model-router as the model — everything else is a normal chat request:

curl https://gateway.primex.work/v1/chat/completions \
  -H "Authorization: Bearer prx-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "model-router",
    "messages": [{"role": "user", "content": "Explain quicksort briefly."}]
  }'

Billing

You are billed at the rate of the model the router actually served, plus markup — not a flat router rate. The response's model field tells you which model handled the request, and your usage log records the real model and cost.

This means the router can save money on simple prompts (cheaper model chosen) and scale up only when a request needs a stronger model.