# Methodology

> Markdown mirror of https://router.one/methodology for AI assistants and crawlers. Router One is an OpenAI-compatible LLM API gateway.
> Last updated: 2026-09-05

Every number Router One publishes traces to one of seven documented sources: the China latency benchmark (p50 measured from Beijing, Shanghai and Shenzhen), routing methodology (EWMA latency scoring, same-model failover), pricing methodology (posted per-model rates, no hidden token markup; FX and channel fees shown at checkout), data retention (direct API metadata, saved Playground history, and deletion requests), security, availability & SLA, and the status page. The pages below document how Router One measures, prices, retains, and stands behind the claims you see across the site.

## [China latency benchmark](https://router.one/benchmarks/china-latency)

p50 and timeout rate across Beijing, Shanghai, Shenzhen — 7-day rolling window, three ISPs.

## [Smart routing](https://router.one/routing-methodology)

Default model_name routing, model:auto server-side candidates, optional latency/cost/reliability signals with time-decayed EWMA scoring, and same-model provider fallback.

## [Pricing & rate updates](https://router.one/pricing-methodology)

No hidden markup on the pay-as-you-go token line. Wallet billing and subscription plans are live. RMB conversion and channel fees are shown at checkout.

## [Data retention](https://router.one/data-retention)

Direct API calls retain metadata, not prompt or reply bodies. Playground saves conversation history. Confirm retention and deletion scope with support.

## [Security](https://router.one/security)

Transport, key handling, upstream processing, and the scope of request-data capture.

## [Availability & SLA](https://router.one/sla)

Uptime measurement, automatic provider fallback, enterprise SLA with credits.

## [Status](https://router.one/status)

Public incident state, covered components, and rolling 90-day status history.

Questions about how something is measured? Email us via the [/contact](https://router.one/contact) page.

## See also

- Canonical page: https://router.one/methodology
- Models and per-model token rates: https://router.one/models (markdown: https://router.one/models.md)
- Pricing: https://router.one/pricing
- API docs (markdown): https://router.one/docs.md
- Company facts: https://router.one/facts/company.md