Sebastian Raschka has released eight videos totaling more than eleven hours on implementing reasoning models. Starting from pretrained Qwen3 0.6B, the course builds mathematical answer verification, self-consistency, self-refinement, GRPO reinforcement learning and distillation in Python and PyTorch. It does not pretrain a base model from scratch. Official code and setup guidance accompany the lessons; early generation and evaluation can run on a CPU, while a GPU is more practical for later training experiments. Familiarity with Python and PyTorch is recommended.