Text this: Evaluating DL Model Scaling Trade-Offs During Inference via an Empirical Benchmark Analysis