Hacker News Reader

Do anyone tried any harnesses benchmark? It would be interesting to compare different harness impact on results from the same model

Artificial Analysis is working on it. They have had some of this, but are aiming for something more thorough and dedicated.