10x Cheaper Inference Path

Friedberg highlights new MIT pruning research suggesting 90% model size reduction with equal accuracy, potentially cutting inference costs by 10x.