search

withmartian / routerbench

The code for the paper ROUTERBENCH: A Benchmark for Multi-LLM Routing System
MIT License
87 stars 17 forks source link

how do we add new models #7

Closed pulkitmehtaworkmetacube closed 3 months ago

pulkitmehtaworkmetacube commented 4 months ago

Please suggest how do we add new models like GPT4-O , Gemini etc in models to route . What is process to create eval data ?

onjas-buidl commented 3 months ago

This repo is LLM-agnostic. If you have new LLMs, you just produce their data in the specified format and you can convert them into the standardized format, and the evaluation pipeline will work accordingly.