Researcher: OpenAI can't stop distillation at medium reasoning effort

adam_karvonen · x · 2026-11-22

Researcher Adam Karvonen argues OpenAI's anti-distillation blog post is a big deal: at medium reasoning effort (what most users run), OpenAI apparently can do nothing to prevent distillation and protect its product; it can only restrict very high reasoning, which is rarely used. The attached SEC-bench Pro chart shows exact reasoning traces distilled after RL scale far better with inference budget (38% vs 13%) than reasoning summaries after fine-tuning — underscoring how little the protections do against summary-based distillation.

Related event: Researcher Says OpenAI's Mid-Tier Reasoning Can't Prevent Distillation(2 posts)→

Original post →

More from Models

Models channel →