UX Improvement: Leaderboard that compares parameter size

[TomLucidor]

Got this idea from EvalPlus and BigCodeBench, and that sometimes it would be good to do apples-to-apples between models, and that if most of the top models are large or proprietary, it does not mean much when it is less accessible to the masses (besides the problem that large models are scalable models). Newer versions of Qwen2 and Phi-3 comes to mind.