An independent research project by Logan.
The results here come from my own experiments using open models, public APIs, local hardware, and rented GPUs.
The Trust Table also uses our own API account. Claude models play and are scored here. We never train on their outputs.