State-of-the-art AI shouldn't cost a fortune.
Botscanner is where you use the router. It sends every prompt to the model that answers it best for the least money. You save over 90% against always calling the top frontier model.
Powered by AutoBench. Current numbers live on autobench.org.
We benchmarked commercial LLM routers on AutoBench. Most of them collapse onto a single favourite model. None of them beat simply always calling the best model in the pool.
So we trained our own on AutoBench data. That is the Botscanner Router (beta).
You ask. The router decides. You see which model answered, and what it cost. A lab cannot honestly route traffic away from its own models. We do not build those models. We measure them, then pick.
Use it in chat, or as an OpenAI-compatible API at https://api.autobench.org/v1. We update the router and the models behind it as better and cheaper models appear.
The Playground is for when you want to see the answers, not just the winner. Send one prompt to several LLMs at once. Other LLMs read those answers and rank them, best first, so you can compare how each model actually responded. The router is for when you want that choice made for you.
AutoBench is the evaluation layer under the router. A committee of LLMs scores each answer — Collective-LLM-as-a-Judge — so the measurement is not one lab grading its own model. Those scores are what the router is trained on. Public runs have evaluated 120+ models and produced 1.3M+ judgments.
The benchmark methodology was independently validated with DIAG, La Sapienza University of Rome (arXiv:2510.22593).
Botscanner is a service of AutoBench, a project of eZecute S.R.L., Rome. The same app is available at botscanner.ai and botscanner.autobench.org.
Sign up with your invitation code and start on the Router with your free credit, or request an invitation code.