Illustration, AI-generated from the story text. Tap a panel to read the story through it.

Route Anything tested model and policy routing across 25 and 374,401 scored runs.

Choosing the best candidate per input beats the best fixed choice by 5.4 points, while learned routers managed 0.3 across 16 benchmarks, though they did better on video and robot tasks.

Products that sell automatic model selection should be asked which of those two numbers they deliver.
A perfect router would beat any single model by 5.4 points, and learned routers capture only 0.3
Route Anything tested model and policy routing across 25 benchmarks and 374,401 scored runs. Choosing the best candidate per input beats the best fixed choice by 5.4 points, while learned routers managed 0.3 across 16 benchmarks, though they did better on video and robot tasks. Products that sell automatic model selection should be asked which of those two numbers they deliver. SourcesA
Sources
- ARoute Anything (Hugging Face Papers)huggingface.co
This story ran in the Afternoon Brief of November 4th 2026.