After the massive leap in national model development a few months ago, progress has slowed down. The reason? Both OpenAI and Anthropic have implemented anti-distillation measures, making the cost and difficulty of distillation 10 times higher than before. Take OpenAI for example—GPT-6 no longer directly displays chain-of-thought reasoning, forcing distillers to use more rounds of questioning to reconstruct GPT's thought process. What used to take 2-5 rounds of interaction now might require 10 or even 30 rounds, driving API call costs up by more than 5x. Anthropic's new protective measures have reduced reasoning token counts by 90–99%. This includes restricting the transfer of reasoning records across conversations, models, and accounts, deploying two-stage classifiers to detect and block extraction requests, and throttling or banning accounts that repeatedly attempt to bypass these safeguards.