Seems interesting but testing myself only yields my results? How would I compare the result to a frontier model, that part seems to be missing?

Also, the tests seem to be heavily skewed in favor of what LLMs are good at.