Florian Brand discusses the narrowing performance gap between open and closed models. The conversation focuses on Kimi K3, Qwen 3.8, and the technical nuances of model distillation. These insights reveal how smaller models inherit capabilities from larger ones. Practitioners can use these patterns to optimize local deployments without sacrificing reasoning quality.