Most comparisons of online math courses list platforms and their prices. That misses the distinction that actually determines whether you learn anything: some of these are courses, and some are video libraries with a course-shaped marketing page. The difference is whether you spend most of your time watching or most of your time stuck.
Here is what each option is genuinely good at, by level.
The distinction that matters
A course has sequenced material, problems you must attempt, and feedback on whether you were right. A library has excellent explanations and no mechanism to make you produce anything. Libraries are wonderful and they are not sufficient. The research on this is unambiguous and slightly humiliating: watching a worked solution produces a strong feeling of competence that does not survive contact with a blank page.
Pick one course as your spine. Use libraries as the explanation layer when the spine loses you.
School level, catching up, or rebuilding foundations
Khan Academy is the correct default and it is free. Arithmetic through algebra, geometry, trigonometry, precalculus, differential and integral calculus, and introductory statistics, each with a substantial exercise bank, hints, and mastery tracking that will not let you skip the thing you cannot do. Its weaknesses are real: the explanations are efficient rather than beautiful, and the problems are mostly routine rather than the multi-step kind that appears on hard exams. Neither is a reason to start elsewhere.
IXL and DeltaMath are practice engines rather than courses. DeltaMath in particular is what many US high school teachers actually assign, and it is strong on generating unlimited variants of a specific problem type.
Art of Problem Solving is the opposite of Khan in every way: hard, discovery-based, aimed at competition and at students who are bored by routine. If a student finds school math too easy, this is the answer. If they are struggling, it will make things worse.
Calculus, linear algebra, and the standard university sequence
MIT OpenCourseWare is the strongest free option at this level. Full lecture courses тАФ 18.01 single-variable calculus, 18.06 linear algebra тАФ with problem sets, exams and solutions. Gilbert Strang's linear algebra course is still, decades on, the best introduction to the subject that exists in any medium. What you do not get is any help when you are stuck, which is the entire difficulty of self-study.
Paul's Online Math Notes is a free, plain, complete set of notes and practice for algebra, calculus I through III and differential equations. Undramatic and better than most textbooks for the specific job of "I need to understand this section tonight."
Coursera and edX carry real university courses, often the same content the institution teaches. What you are paying for is graded assignments, deadlines and a certificate. Audit tracks are free on many courses and give you the material without the grading. Be aware that quality varies enormously between courses on the same platform.
OpenStax provides free, peer-reviewed, genuinely good textbooks for precalculus, calculus and statistics, with instructor answer keys. If you want a textbook, take this one instead of buying anything.
Intuition and motivation
3Blue1Brown is not a course and does not pretend to be. It is the best conceptual explanation of linear algebra, calculus and neural networks available, and it works best alongside a course, not instead of one. Watch the essence-of-calculus series before your first calculus term and again in week ten.
Brilliant sits between the categories: interactive, puzzle-driven, subscription-priced, excellent for building intuition and for adults returning to math who want something enjoyable. It is not a curriculum and will not prepare you for a specific exam.
Mathigon is free, interactive and unusually beautiful, strongest at school-level geometry and number topics.
For adults returning to math
The mistake is starting where you think you should be. Nearly everyone rebuilding math as an adult stalls not on the new material but on fractions, negative signs, exponent rules and algebraic manipulation тАФ the machinery below the topic they are trying to learn. Take a Khan Academy diagnostic and start where it puts you, even if that feels embarrassing. Two months on foundations saves a year of frustration.
Where a solver fits тАФ and where it does not
Self-study has one structural weakness: no feedback when you are stuck. That is the gap AI tools fill, and it is worth being precise about what fills it and what does not.
A solver that shows worked steps answers "why is my answer different from the book's" тАФ the single most common self-study blocker. Working through a limit, an integral or a linear equation step by step lets you locate the exact line where your work diverged. That is a genuinely useful loop, and it is what AI-Math is for.
What a solver does not do is teach you a syllabus, decide what to study next, or make you attempt anything. It has no idea what course you are taking. Use it after you have attempted a problem and got a different answer, never as the first move тАФ that inverts the learning and you will feel productive while retaining nothing.
A working stack
- Rebuilding foundations: Khan Academy as the spine, 3Blue1Brown for intuition, a solver for stuck moments.
- High school, aiming high: school course plus Art of Problem Solving, DeltaMath for volume.
- University calculus or linear algebra: your actual course, MIT OCW as the second explanation, Paul's Notes as the reference, OpenStax as the textbook.
- Adult, curiosity-driven, no exam: Brilliant plus 3Blue1Brown, with Khan for anything that turns out to be missing.
Almost all of that is free. The thing that is scarce is not material тАФ it is the hours you spend with a pen, stuck, before you look anything up.
Related reading: