Sebastian Raschka pubblica un corso gratuito per costruire modelli di ragionamento AI da zero

What you could do with it

• A machine learning researcher could teach themselves evaluation, inference-time scaling, reinforcement learning, and distillation by coding along with Raschka's eight videos in Python and PyTorch ¹.
• An individual with some coding experience could use the course to understand how current AI reasoning systems think and solve problems by building on a pre-trained Qwen3 model ¹.
• A student could read a chapter from Raschka's companion book and then watch the matching video before trying the exercises for a complete learning experience ¹.

What the course covers

The course builds on a small pre-trained Qwen3 language model and walks through evaluation, inference-time scaling, reinforcement learning, and distillation — all implemented in Python and PyTorch ¹. Raschka says this approach is still the best way to understand how these systems really work .

The eight videos total over 11 hours of content and assume you are comfortable with Python and PyTorch ¹. Some early examples can run on a CPU, but later training experiments will run much better with a GPU .

How reasoning models fit into AI agents

Raschka also explains how reasoning models relate to AI agents — the software frameworks that let models do things like code, book travel, or manage your calendar ¹. In his analogy, the reasoning model is the engine, and the agent harness is the rest of the car that lets you use it for real tasks . Most current generation LLMs already include a reasoning mode, though they vary in how much "thinking effort" they apply .

How to follow along

The videos are meant to go with Raschka's book Build a Reasoning Model (From Scratch), available from Manning and Amazon ¹. He recommends reading a chapter, watching the matching video, then doing the exercises . For a quicker overview, he also points to a 90-minute LinkedIn Learning course on the same topic .

His earlier course, Build a Large Language Model (From Scratch), is helpful background but not required ¹.

Who is Sebastian Raschka?

Raschka is a machine learning researcher and author who has been writing about reasoning models since at least two years ago ¹. He previously published the popular Build a Large Language Model (From Scratch) book and course, which covered the fundamentals of how large language models work under the hood .

Raschka shared the course announcement in a post on his Substack blog, Ahead of AI ¹.

This story reports on a trend for information only. It is not advice or encouragement to do any of it.

I fatti principali

• 8 videos, over 11 hours of coding
• Free to access
• Built on a small pre-trained Qwen3 model
• Covers evaluation, inference scaling, reinforcement learning, and distillation
• Requires Python and PyTorch; GPU recommended for later experiments

Fonti

1. Build Your Own Reasoning Model: A Full Video Course

Guarda