Routing & fallbacks

Each model slug can be served by more than one upstream provider. The gateway picks a healthy, competitively priced provider automatically, so a single provider outage does not take your app down.

Automatic retries

If the chosen upstream errors or times out, the request is retried against another provider for the same model. You are only billed for successful generations.

Model fallbacks

Pass a models array instead of a single model to fall back across models in priority order — useful for keeping a premium model first with a cheaper safety net behind it.

JSON
{
  "models": [
    "moonshotai/kimi-k3",
    "deepseek/deepseek-v4-pro",
    "qwen/qwen3.8-max"
  ],
  "messages": [{"role": "user", "content": "Hello"}]
}

Latency

The gateway is hosted in Japan with anycast routing, keeping first-hop latency low for traffic across East and Southeast Asia.