Benchmarks are the BMI of model evaluation. They may have utility in trying to look at the whole landscape of models, but are very misleading when it comes to

Benchmarks are the BMI of model evaluation. They may have utility in trying to look at the whole landscape of models, but are very misleading when it comes to making 1:1 comparisons or in developing confidence at to how a given model will deliver on your workflow.