Which Offerings of Stanford CS336 Are Available?

Stanford CS336 currently lists Spring 2026 as the active offering, with Spring 2025 and Spring 2024 available as previous offerings. If you want to follow the course now, use the Spring 2026 schedule and materials. If you are comparing changes over time or looking for older assignments and recordings, the previous offerings are the relevant reference points.

Current Offering: Spring 2026

The official course site identifies CS336: Language Modeling from Scratch as a Stanford / Spring 2026 course. The page also links to previous offerings directly from the title area.

Key logistics for Spring 2026:

Course staff listed for the offering:

• Tatsunori Hashimoto, Instructor
• Percy Liang, Instructor
• Herman Brunborg, CA
• Marcel Rød, CA
• Steven Cao, CA

Office hours listed on the page:

• Percy Liang: Fridays 11am–12pm, Gates 366
• Tatsu Hashimoto: Tuesdays 11–12am, Gates 364
• Marcel Rød: Tuesdays 4:30–5:30pm, Gates 498; Wednesdays 4:30–5:30pm, Gates 415
• Herman Brunborg: Wednesdays 1:30–2:30pm, Fridays 1:30–2:30pm, Gates 392
• Steven Cao: Mondays 4:30–5:30pm, Thursdays 9:30–10:30am, Gates 200

The page notes that students should ask all course-related questions in public Slack channels, and that all announcements will also be made in Slack.

Previous Offerings: Spring 2025 and Spring 2024

The site explicitly lists Spring 2025 and Spring 2024 as previous offerings. These are useful if you want to:

• Compare how the course has evolved across terms
• Review older schedules or assignment structures
• Use prior recordings or materials when the current term’s content is not yet available

The page does not provide separate logistics tables for Spring 2025 or Spring 2024 in the excerpt. To compare those terms in detail, you would need to open the linked previous offerings from the course site.

What the Course Covers

The course walks through the entire process of developing a language model from scratch, including:

• Data collection and cleaning for pre-training
• Transformer model construction
• Model training
• Evaluation before deployment

The course description compares its approach to operating systems courses that build an entire OS from scratch. Assignments are primarily in Python, and the page states that students receive minimal scaffolding. The amount of code written is described as at least an order of magnitude greater than in other AI classes.

Prerequisites to Check Before Choosing an Offering

The page lists two main prerequisite areas:

1. Proficiency in Python — Most assignments are in Python, and the course expects strong software engineering ability.
2. Experience with deep learning and systems optimization — A significant part involves making neural language models run quickly and efficiently on GPUs across multiple machines. Students are expected to know PyTorch well and understand basic systems concepts such as the memory hierarchy.

If you are deciding which offering to follow, the prerequisites are the same across terms as far as the page indicates. The main difference is access to current lectures, office hours, and Slack versus archived materials from Spring 2025 or Spring 2024.

How to Choose

• Follow the course live or need current logistics: Use Spring 2026.
• Want to study at your own pace or compare content over time: Start with Spring 2025 or Spring 2024.
• Need recordings: The Spring 2026 page links to a YouTube playlist. Previous offerings may also have recordings, but the excerpt does not confirm their availability.
• Need to contact staff: Use Slack for course-related questions and the listed email for personal matters. This applies to the Spring 2026 offering.

The course site is the authoritative source for schedule, assignments, and materials. Check the previous offerings links if you need term-specific details beyond what the current page summarizes.