MakoBench

LLM benchmark board: open-weights models on our own GPUs, hosted models through their makers' APIs.