Build a Reasoning Model (From Scratch)

In this book, Sebastian Raschka guides readers through creating a working reasoning model from scratch, starting with a pre-trained LLM and implementing core reasoning improvements.

You'll learn to evaluate models with judgment- and benchmark-based methods, enhance reasoning without weight updates, integrate tools via reinforcement learning, apply distillation from larger models, and master the full development pipeline.

Reasoning models excel in math, logic, and code by breaking problems into steps, mirroring techniques in top models like Grok 4 and GPT-5.