Your first request

Send a request, read the routing headers, iterate.

Once you have a funded account and an API key, the fastest way to feel the router is a one-line curl against the chat-completions endpoint.

bash

curl https://www.hiway2llm.com/v1/chat/completions \
  -H "Authorization: Bearer hwy_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      {"role": "user", "content": "What is the capital of France?"}
    ]
  }'

The response headers tell you what happened

HiWay adds three custom headers on every successful response:

X-HiWay-Model — the exact provider model that answered the request.
X-HiWay-Tier — the routing tier picked (light, standard, heavy).
X-HiWay-Cost-USD — the exact cost of this request, in USD, debited from your balance.

Deterministic routing

The same prompt always routes to the same tier — we use a pure CPU scoring engine, no ML, no randomness. You can replay a request and get the same tier back as long as your routing profile hasn't changed.

What `model: auto` means

Passing auto tells HiWay to pick the best tier for the prompt based on complexity. You can also pin a tier (e.g. light, standard, heavy) or a specific provider model (e.g. claude-sonnet, gpt-4o-mini). Pinning overrides the router and goes straight to the provider.