# Router One Smart Routing & Fallback Facts

Last updated: 2026-06-10

- Product: Router One
- Category: LLM API Gateway / AI Model Router
- Routing signals: EWMA latency over the last 50 requests per provider per model family; 5xx and gateway-timeout rate over the same rolling window; posted token-level cost per candidate; per-project latency/cost/quality weights
- Provider down-ranking: crossing the configured error-rate threshold temporarily down-ranks a provider
- Model selection: pin an exact model for determinism, or send model="auto" to let the gateway pick a candidate using the configured weights
- Fallback triggers: provider 5xx response, network error, or response time exceeding the per-model timeout budget
- Fallback scope: constrained to the same model family (GPT to GPT, Claude to Claude); the exact variant used is recorded in the per-request trace
- Fallback latency: typical end-to-end fallback adds under 200ms over the failing request
- Fallback opt-out: enterprise contracts can disable fallback per project; pay-as-you-go projects use the default fallback configuration
- Trace contents: model, provider, input/output tokens, cost, latency, status code, and the route or fallback decision — including the failed provider, error code, and latency when fallback occurred
- Trust boundary: smart routing selects upstream routes for LLM API requests; Router One is not an agent orchestration runtime or workflow engine

## Canonical routing pages

- Smart routing methodology: https://router.one/routing-methodology
- Automatic LLM fallback: https://router.one/llm-fallback
- LLM API gateway overview: https://router.one/llm-api-gateway (markdown: https://router.one/llm-api-gateway.md)
- China latency benchmark: https://router.one/benchmarks/china-latency
- Company facts: https://router.one/facts/company.md