RouteWorks/RouterArena

The platform evaluates routers across five specialized metrics—accuracy, cost-efficiency, optimal selection rate, robustness, and latency—using a principled dataset spanning 9 domains with difficulty stratification. It provides automated end-to-end evaluation pipelines compatible with both open-source and commercial routers via standardized config files, with extensible support for custom router implementations and multiple LLM inference providers (OpenAI, Anthropic, Gemini, etc.).