LeetCode vs HackerRank vs CodeSignal: which one sends you the assessment link?

Here’s a small experiment. On 20 August 2026 I asked all three pricing pages for their prices using a plain HTTP request, no browser and no cookies. CodeSignal handed over a full plan table. HackerRank bounced me to its work pricing URL and served a cookie notice where the plans should have been. LeetCode returned a 403.

The LeetCode vs HackerRank vs CodeSignal question is really two questions sharing one search box. One is about which site you grind on tonight. The other is about which platform grades you in three weeks. Those have different answers, and the business models are where the split shows up first.

Two of these are sold to employers. One is sold to you.

Open CodeSignal’s pricing page and the first thing it does is ask who you are. There’s a Hire side, a Learn side and an Education side. Hire runs three tiers called Build, Grow and Pro. Learn has a free tier and a paid one called Cosmo+, listed at $24.99 a month, described as unlimited access with an AI tutor and personalized paths. One company, two products, two completely different customers. The Education side lists three plans, Department Pilot, Multi-Program Roll Out and Institution-wide, and all three are quoted by sales rather than priced on the page, as is the Teams and Enterprise option on the Learn side.

HackerRank’s Screen product page doesn’t bother with the split. It reads like a page written for somebody holding a purchase order. Named ATS and HRIS integrations (Greenhouse, Lever, Workday, SAP SuccessFactors, iCIMS). An AI plagiarism detection model. An AI proctor that can intervene during a test. Identity verification and multiple-face monitoring. Auto-scored assessments with structured candidate results. A question library the page puts at 77 roles and more than 260 skills, with custom tests built from a bank of over 7,000 questions. None of that is something a candidate would pay for. All of it is something a hiring manager would.

LeetCode sits on the other side of the table. It sells a subscription to the person doing the practicing, and that has been true for long enough that the brand name has turned into a verb. I can’t print a Premium price here. The subscribe page returned a 403 to my request on 20 August 2026, and reprinting a number I didn’t see with my own eyes is how stale prices spread.

The confusion is understandable, because all three show you the same surface. A problem statement on the left, a code editor on the right, a run button. Identical furniture, different rooms.

Which platform do employers actually use?

If an automated coding test arrives before any human has spoken to you, it’s coming from a screening product that somebody at the company bought. HackerRank and CodeSignal both sell one. LeetCode doesn’t sell to your employer the same way; it sells practice to you. So the platform grading you is usually not the platform you trained on.

I can’t give you a market share figure. What I can tell you is what each company sells and who signs the invoice, which is a slower route to the same answer and a more honest one.

There’s a second signal buried in CodeSignal’s tier list. Basic ATS integrations appear on the Grow plan, enterprise ATS integrations on Pro, and advanced fraud prevention on Pro as well. So the deep wiring into a company’s hiring stack sits at the top of the price list. That’s roughly what you’d expect, and it also means a five-person startup sending you a CodeSignal link is probably running the process by hand rather than through an automated pipeline.

Here’s how LeetCode vs HackerRank vs CodeSignal splits on the four things I could check from the vendors’ own pages on 20 August 2026.

What LeetCode vs HackerRank vs CodeSignal costs, and who pays

Every figure below came off the vendor’s own page on 20 August 2026. Where a cell says the price wasn’t shown, that means exactly that. The page didn’t render one for me, so I left it empty rather than fill it in from memory or from somebody else’s blog post.

Two things about that table are worth saying out loud. The first is that a candidate can see what a candidate pays on exactly one of the three sites. The second is that CodeSignal is charging both sides of the same hiring transaction, which I find slightly odd. The company selling your employer a screening test also sells you the tutor that prepares you for it.

The credit math, and the bit I couldn’t check

CodeSignal’s Hire plans are metered in credits, and the two published tiers are internally consistent in a way worth noticing. Build lists 5 credits a month or 60 a year. Grow lists 35 a month or 420 a year. Both annual figures are exactly twelve times the monthly one, which means the annual discount is applied to the price and not to the volume. The prices agree with that reading: $79 against $99, and $479 against $599, are both a fifth off, matching the 20 percent annual saving the page advertises.

Divide it out and Build lands near $15.80 per credit on annual billing, Grow near $13.69. Grow costs about six times more per month and carries seven times the credits. That’s the usual shape of a metered product, so no surprise, but it does mean the jump from Build to Grow is the point where a company stops rationing tests.

What I couldn’t check is what one credit actually buys. The pricing page lists credits per plan and I didn’t verify whether a credit equals one candidate, one assessment, or one interview session. Treat my per-credit division as arithmetic on a unit I can’t define. Don’t build a hiring budget on it, and if you’re on the buying side, ask.

What the assessment link tells you before you open it

Once you know who paid, the link in your inbox reads differently.

A CodeSignal test costs the company money per candidate. Metering changes behavior in predictable ways: a team on Build with five credits a month isn’t spraying assessments across everyone who applied, and they’ve very likely picked a passing bar before you start rather than after they see the distribution. That’s an inference from the pricing model, not something CodeSignal told me, so hold it loosely.

A HackerRank test tends to arrive with the integrity stack switched on. Their Screen page lists an AI proctor that can intervene, identity verification, multiple-face monitoring and a plagiarism detection model. The practical translation is short. Sit alone. Keep one face in frame. Don’t paste in code you didn’t type in that window. Assume something reviews the session after you close the tab, because that is exactly what the vendor is selling to the person who sent you the link.

The HackerRank pitch has also moved, and this is the part I’d pay attention to if I were preparing this month. The page frames the whole product around hiring for AI-first work, and names three things it says it measures: fundamentals without AI, the ability to work with AI across the development cycle, and knowledge of AI concepts like prompt engineering, RAG and vector databases. Whatever you make of that framing, it describes a different test from the one most people spent 2023 drilling for.

LeetCode has none of that machinery pointed at you. No proctor, no identity check, no score travelling anywhere. Which is the point of a gym, and also the reason a strong LeetCode streak doesn’t transfer as cleanly as people expect.

How I’d prepare for each one

For the algorithm screen, patterns beat volume, and that’s been true long enough that I’d call it settled rather than debatable. Learn the dozen or so shapes that keep recurring instead of solving 400 problems and hoping the mapping happens by itself. Our LeetCode patterns cheat sheet lays the recurring ones out, and the Blind 75 guide is the shorter runway if the screen is next week.

For a HackerRank or CodeSignal screen, practice the format and not only the problem. A visible countdown. Input parsing that isn’t handed to you as a tidy function argument. Hidden tests that pass or fail without explaining themselves. One submission, scored by a machine, with no interviewer in the room to notice that your approach was right and your off-by-one wasn’t. Those constraints break people who solve the identical problem fine in their own editor at their own pace, and the fix is boring: do a few full-length timed runs on the actual platform before the real one.

For the live rounds that come after the screen, the prep is a different discipline again, since a human is asking follow-up questions rather than a grader counting passing tests. That’s the gap LastRoundAI’s interview copilot was built for, and I’d keep it mentally separate from problem grinding rather than treating both as the same preparation.

The part I’d argue about

I think the algorithm grind is currently mispriced, and I might be wrong about it. HackerRank is selling employers a test of whether you can work with AI, not only whether you can reverse a linked list under time pressure. If that pitch lands with buyers over the next few hiring cycles, the skill people spent two years drilling becomes the smaller half of the screen rather than the whole thing.

The counterargument is decent, and I hold it at the same time. The test a mid-size company sends you next Tuesday may still be two array problems and a forty-five minute timer. I don’t have data on what fraction of live HackerRank sessions use the newer question types, and I did look. If you have that number, I’d like to see it.

One more caveat, since this is the kind of post people quote back six months later. My read of who buys what rests on four pages across three vendor sites, all opened on a single day, three of which I’ve linked above so you can check them yourself. Pricing pages change quietly. Look at the date stamp on anything you read about this, including this.

The way I’d frame LeetCode vs HackerRank vs CodeSignal now: you practice on one and get graded on another, and the two don’t have to agree with each other. The cheapest move available to you costs one email. Ask the recruiter which platform the screen runs on, before you spend a week preparing for the wrong format. Most of them will just tell you.