The verdict in 60 seconds
UC Berkeley is, quietly, the best university in the world for free computer science self-study — not through a MOOC platform, but because its flagship courses publish nearly everything on public course sites: lectures, slides, homework and the full projects. CS61A (cs61a.org), CS61B, Data 8 (data8.org), CS188 and CS285 between them cover a serious chunk of a CS degree for £0. The catch: no certificate, no credit, no help desk, and current-semester autograders are usually gated to enrolled students. For certificated study, BerkeleyX on edX offers free audit tracks with paid certificates around £100–£140 — The Science of Happiness being the famous one. The old lecture webcast archive was deleted in 2015, so ignore any guide pointing you at dead YouTube playlists; use the live course sites below.
Most “free Berkeley courses” articles point you at edX and stop. That misses the real story. Berkeley’s computer science and data science departments run their courses on open websites as a matter of habit, which means the actual courses Berkeley undergraduates take — the same projects, the same assignments, roughly the same lectures — are sitting in public. Nobody markets this, there is no enrolment funnel, and that is precisely why it is good. The trade-off is that you are on your own: no certificate exists for any of it, semester sites reorganise every few months, and some graded infrastructure is closed to outsiders. This guide lists what is genuinely available in 2026, who each course actually suits, how long each one takes when you do the projects properly (the projects are the point), and where the paid edX certificates fit if you need paper. It is a long list; nobody should attempt all of it. Pick one course, finish it, then come back.
How we chose what made this list
Four criteria. First, free to complete in full: lectures or detailed notes, assignments, and projects all publicly accessible — not a syllabus page with dead links. Second, the real Berkeley course or an official Berkeley MOOC: everything here is either an actual course site run by Berkeley faculty or a BerkeleyX course on edX, not third-party summaries. Third, working in 2026: we checked that the current or most recent public offering loads and that materials are complete enough to self-study from. Fourth, honest labour estimates: Berkeley courses are famously heavy, so we give realistic hour counts based on the workload enrolled students report, not the marketing figure. Where a paid certificate exists we give the price in pounds; where none exists we say so plainly.
The best free Berkeley courses in 2026
1. CS61A: Structure and Interpretation of Computer Programs
Berkeley’s legendary first course, taught for years by John DeNero, with everything public at cs61a.org: lecture videos, slides, labs, homework and the three big projects (the Hog dice game, the Ants tower-defence game, and building a Scheme interpreter). It teaches programming through Python, then Scheme and SQL, with a focus on abstraction rather than syntax. Best for people who have written a little code and want to think like a computer scientist; brutal as a literal first exposure to programming. Expect 150–200 hours done honestly. No certificate, no credit. Cost: £0. If any single free CS course deserves the word “complete”, it is this one.
2. CS61B: Data Structures
The second course in the sequence, in Java. Josh Hug’s Spring 2021 offering was built deliberately for outside learners — public lectures, homework, and the famous Gitlet project, where you implement a working version of Git from scratch. That project alone teaches more software engineering than most paid bootcamps. For self-learners who have finished CS61A or equivalent and can already program. The public offering includes autograder access for most assignments, which is rare and valuable. Expect 150–200 hours. No certificate. Cost: £0. Weakness: the sp21 materials are now five years old, though data structures have not changed.
3. Data 8: The Foundations of Data Science
Berkeley’s gentle, genuinely beginner-friendly introduction to data science at data8.org, built for students with no programming background. It teaches Python, statistical inference and prediction together, and the accompanying textbook, Computational and Inferential Thinking, is free online in full. This is the right starting point if CS61A sounds intimidating or your interest is data rather than software. Materials, labs and projects are public; you can run everything in a free hosted notebook. Around 80–120 hours. No certificate from the course site (a related BerkeleyX Data Science programme on edX offers paid certificates, roughly £120 per course). Cost: £0.
4. CS188: Introduction to Artificial Intelligence
The classic AI course, home of the Pacman projects — you implement search, minimax, reinforcement learning and probabilistic inference to drive Pacman around mazes. Lectures, notes and the project code are public via the course site and ai.berkeley.edu, and the Pacman assignments are used by universities worldwide because they are that good. Best after CS61A-level programming plus some probability. It covers classical AI thoroughly; know that modern deep learning gets limited treatment, so pair it with something newer if that is your goal. Around 100–150 hours. No certificate. Cost: £0.
5. CS285: Deep Reinforcement Learning
Sergey Levine’s graduate course, with full lecture videos on YouTube and assignments on GitHub. This is one of the most-watched RL courses in the world, and it is the real thing: policy gradients, Q-learning, model-based RL, offline RL, taught by one of the field’s leading researchers. It is for people with machine learning, linear algebra and PyTorch already in hand — as a first ML course it will flatten you. The homework requires a capable GPU or cloud credits (budget £20–£50 if you rent). 100+ hours. No certificate. Cost: £0 plus compute.
6. CS61C: Great Ideas in Computer Architecture
The third leg of the lower-division sequence: C, assembly, how memory actually works, caches, parallelism, building a CPU in a logic simulator. Recent semester sites publish lectures, labs and projects publicly. It is the least glamorous course on this list and possibly the most differentiating — self-taught developers routinely skip architecture and it shows in interviews and debugging. Take it after CS61B. Around 120–150 hours. No certificate. Cost: £0. Weakness: project infrastructure (Venus, Logisim setups) occasionally assumes campus tooling, so expect some environment fiddling.
7. The Science of Happiness (GG101x on edX)
The one non-CS entry, because it is Berkeley’s most successful online course ever: positive psychology from the Greater Good Science Center, with over 850,000 enrolments since 2014. The audit track is free for the duration of the run; the verified certificate costs about £130 ($169). The content is research-grounded — this is a psychology course with weekly practices, not a motivational product — and it suits anyone, no prerequisites. Eight to eleven weeks at 2–3 hours per week. Honest note: audit access expires, so finish within the run or pay; and a certificate here is a personal-interest credential, not a professional one.
8. Agile Development Using Ruby on Rails (CS169.1x on edX)
Armando Fox and David Patterson’s software engineering course, the edX version of Berkeley’s CS169. It teaches SaaS architecture, test-driven development, and agile practice with real tooling, and the accompanying textbook is one of the better software engineering texts available. Audit is free; the verified certificate runs to roughly £110. Ruby on Rails is no longer fashionable, which matters less than you would think — the engineering practices transfer directly — but be aware you are learning a 2010s stack. For programmers who can already build small things and want engineering discipline. 60–100 hours across the sequence. Cost: £0 to audit.
Common mistakes to avoid
1. Hunting for the old webcast archive. Berkeley removed its massive public lecture archive from YouTube and iTunes U in 2015, and countless guides still link to it. If a Berkeley lecture link is dead, do not dig for mirrors of decade-old recordings — the current course sites above have newer material, properly maintained.
2. Watching lectures and skipping projects. A Berkeley course is its projects. Watching CS61B videos without building Gitlet gives you perhaps 20% of the value while feeling like 80%. If time is short, do fewer courses with all the projects rather than more courses without them.
3. Starting in the wrong place. CS285 with no ML background, or CS61B before you can program, ends in abandonment. The honest ordering: Data 8 or CS61A first depending on whether your goal is data or software, then CS61B, then branch to 61C, 188 or beyond.
4. Expecting credit or a certificate from the course sites. The open course sites issue nothing — no certificate, no transcript, no proof. Only the edX courses sell certificates (£100–£140), and none of it is UC Berkeley degree credit; formal credit routes like UC Berkeley Extension cost hundreds of pounds per unit. Build a portfolio of finished projects instead; for these courses, that is worth more than any PDF.
Frequently asked questions
Are these the same courses Berkeley students take?
Substantially, yes. The public course sites are the actual sites enrolled students use, so the lectures, assignments and projects are the real thing. What you do not get: exams, graded feedback, office hours, discussion forums with staff, and in some semesters the current autograder. CS61B’s sp21 public offering and CS61A are the most complete for outsiders; other courses vary semester to semester in how much grading infrastructure is open.
Can I get a certificate for CS61A or CS61B?
No. The open course sites offer no certificate at any price, and no third party can legitimately sell you one. If you need a credential for visa, employer-reimbursement or CV-screening purposes, use the BerkeleyX edX courses (certificates around £100–£140) or a different provider entirely — our Coursera guide covers the certificated equivalents. For software jobs specifically, a finished Gitlet or Scheme interpreter in your GitHub speaks louder than a certificate in practice.
Is Data 8 or CS61A the better starting point?
Data 8 if you have never programmed or your goal is data analysis, statistics or research; it was designed for absolute beginners and its free textbook is exceptional. CS61A if you have written some code and your goal is software engineering; it is harder but builds deeper programming foundations. Doing Data 8 first and CS61A second is a legitimate and common path — the overlap is small and the two courses teach different subsets of Python deliberately.
Do I need a powerful computer?
For everything except CS285, no — any laptop from the last eight years handles CS61A, CS61B, Data 8 and CS188 comfortably, and Data 8 can run entirely in a browser notebook. CS285’s deep RL homework genuinely needs a GPU; a free Colab tier gets you partway, and £20–£50 of cloud credit covers the rest. Do not buy hardware for any course on this list before you have finished its first project on what you own.
What happened to Berkeley’s free YouTube lectures?
In 2015–2017 Berkeley removed roughly 20,000 recorded lectures from public platforms following a US Department of Justice accessibility ruling, since captioning the archive was ruled unaffordable. Some recordings survive in unofficial mirrors of uneven quality and age. Our advice is to let them go: the current course sites listed above publish newer, better-produced material, and a 2012 recording of a course that has since been redesigned mostly wastes your time.
Related guides
If you want certificates and structure rather than raw course sites, start with our Coursera guide. For cheap topic-specific fill-ins alongside these courses, see our Udemy review. And if you want a fully free, structured, project-based path with a supportive community, read our freeCodeCamp 2026 review — it pairs well with CS61A rather than replacing it.