Model Router Benchmarks
By Steven Van ·
The Router Index scores each router from 0 to 10, weighting quality 60%, speed 20% and cost 20% by default, with an adjustable slider.
OpenRouter now publishes Model Router Benchmarks, a page that compares model routers side by side across six suites. Each router is scored on quality, speed and cost, with individual models shown alongside as a baseline: ones representing best-in-class cost efficiency and intelligence.
The announcement sorts routers into types. Provider routing picks among inference providers and falls back if one fails. Model routing sends a request to a router that decides which model answers. The benchmarked model routers fall into three groups:
- Blend of models: Unbiased's Pareto and Sakana's Fugu. They don't disclose which models are used or how often switching happens, and they bill at standardized per-token rates.
- Selects a model per turn: Auto Router and Jev Router. Selection is transparent, and billing is at the standard rate of the chosen model.
- Swaps between a pre-defined set: NVIDIA's Switchyard pairs a cheaper model with a stronger one. Several pairs were benchmarked, but any pairing can be used.
Pareto Code, the -latest slugs and Fusion were not benchmarked, because OpenRouter says they serve specialised purposes that need different comparison methods.
The Router Index combines quality, speed and cost into a 0 to 10 score for each benchmark. The default weights are 60% quality, 20% time per task and 20% cost, and a slider on the page adjusts them. OpenRouter notes the results reflect general tasks, not any one person's work. It says it will keep adding routers and updating scores. All the listed routers are available on OpenRouter and can be swapped in for a model in any app, harness or through the API.
