RelayRank published an open GitHub checklist for testing AI relay stability: https://github.com/wuyabinde223/relayrank-checklist. It helps users move from landing-page claims to reproducible checks for availability, latency, model compatibility, billing behavior, and backup routing.
How to use it
- Start with RelayRank rankings to shortlist providers: https://www.zhongzhuanbang.com/en-US/rankings/
- Run the checklist against your own base URL and API key.
- Record failures, latency, model names, and long-task behavior before increasing budget.
Best fit
The checklist is most useful for Claude Code, Codex, Cursor, and automation workflows where one successful request is not enough to prove stability.