A curated LLM eval package with a benchmark
A curated eval set and an open benchmark: find the model, effort and cache setting that clears your bar at the lowest cost.
A curated eval set and an open benchmark: find the model, effort and cache setting that clears your bar at the lowest cost.