CS224N: Natural Language Processing with Deep Learning

What CS224N is

CS224N: Natural Language Processing with Deep Learning is Stanford's main NLP course. The latest offering on the course site is Winter 2026. Enrolled students get the current lectures through Stanford, while a public YouTube playlist of an earlier year's lectures and the course materials are open to everyone.

What it covers

The course starts with word vectors and neural network basics, then moves through recurrent networks to transformers, pretraining, and post-training methods such as supervised fine-tuning, RLHF and DPO. Later topics include prompting, parameter-efficient fine-tuning, retrieval-augmented generation, reasoning and how to evaluate and benchmark language models. Assignments use PyTorch. The default final project is implementing a minimal GPT-2; students may instead propose their own project on human language.

Is it worth your time?

Yes, if you want to understand why modern language models work, not only how to call them. It gives the reasoning behind each idea, from why word vectors capture meaning to why attention replaced recurrence, which makes you far better at debugging, choosing models and reading papers. Pair it with the Hugging Face course for day-to-day practical skills.

Prerequisites

The course lists proficiency in Python with NumPy and PyTorch, college calculus and linear algebra, and basic probability and statistics, and recommends a foundation in machine learning. If you have finished a deep learning course and can write a PyTorch training loop, you are ready.

How long it takes

About ten weeks at Stanford pace. Self-paced learners should expect longer, especially if they do the assignments, which is strongly recommended.

Free and paid

The public lectures and materials are free. Credit, grading and the current term's recordings are for enrolled students only.

What you can do after it

Explain and implement the core pieces of modern NLP, fine-tune and evaluate language models sensibly, and read current NLP research.

Common questions

Is it only about large language models? No. It covers the foundations, word vectors, neural networks and sequence models, that make large models understandable.

Best for: programmers with some deep learning who want a rigorous, free path into NLP.