Yes, and presumably: more data, more compute, better pre-training and post-training methods, leading to better capabilities.