The best apps for learning Spanish, by what you're actually doing

Every page on this query gives you a numbered list, puts Duolingo at 1, and hands you an "average rating: 4.7/5" scraped off the App Store. It's a ranking of apps against each other, which is a comparison nobody needs.

The question you're actually asking is which one you should install tonight. And I can't answer that without knowing one thing about you, so this page sorts by that instead.

Find the paragraph that describes your week. Read that one. The rest is somebody else's problem.

Beginner, no tutor: Duolingo, or Babbel if you want grammar

This is most of you, and Verbamor is the wrong answer here. I'd rather say that in the first section than bury it.

If you're starting from nothing, download Duolingo. It's free in a way that isn't a trick: the full core Spanish course, ad-supported, gated by Energy, which replaced hearts as the thing that stops you doing 40 lessons in one sitting. Duolingo doesn't publish the numbers behind Energy, so ignore any page that quotes you one.

Two things people repeat about Duolingo that are now out of date, and being wrong about a competitor is a bad look, so let me correct both.

The first is that it has no real spaced repetition. Duolingo published and open-sourced a scheduler: half-life regression, in Settles and Meeder's ACL 2016 paper. That's peer-reviewed work, which is more than most apps on any of these lists can claim. Whether HLR is still what runs in production today isn't something anyone outside Duolingo can verify, so I won't tell you either way.

The second is "Duolingo only gets you to A2." On 22 April 2026 Duolingo said its Spanish course teaches content through CEFR B2, at a Duolingo Score of 129, across 9 courses. That's a statement about what's in the course rather than a measurement of what learners walk away able to do, and the distinction matters. But if a page tells you Duolingo caps at A2 in 2026, that page hasn't been updated in years.

The exception is a specific kind of adult learner: you want the rule stated before you're asked to use it, and inferring grammar from example sentences makes you want to throw the phone. That's Babbel.

Babbel's lessons are short, "most take around 10-15 minutes" by its own description, and its Review feature does use spaced repetition, pulling vocabulary from lessons you've finished and scheduling the sessions for you. Babbel doesn't publish the interval ladder, so I can't tell you how it compares to anything, and neither can any page that claims to.

Price: Babbel sells 1, 3, 6 and 12-month plans plus a lifetime option. The annual works out around $9 a month, though that figure comes from a Babbel coupon page, so treat it as a discounted rate rather than a standing one. Babbel discounts more or less permanently.

One thing to know if you found Babbel while shopping for a teacher: Babbel Live is no longer available to individual consumers. Live teaching means italki, Preply, or someone local.

Never spoken it aloud: buy a tutor, not an app

This one costs me something to write, since I sell an app.

If you've been studying Spanish for a while and have never said a full sentence to another human, no app on this page is your best next purchase. Book an hour on italki.

The reason is that tapping and typing quietly hide the gap. You know the word when it's one of four options on a screen. Out loud, under time pressure, with a person waiting, you find out which words you actually own. One hour does more diagnostic work than six months of solo drilling, and the list of things you couldn't say becomes the best study list you'll ever have.

It's also cheaper than people assume. italki needs no subscription and has no bulk minimum: you buy credits and book. Tutors set their own rates, and there are two tiers, Community Tutors (native or fluent speakers without formal credentials, conversation-focused) and Professional Teachers (formally qualified). On the number: italki's own Spanish tutor page says tutors "cost between $4 to $60 per hour," a band wide enough to tell you nothing. So I read the live listings on 11 August 2026 and copied down the trial prices as they appeared: $5.00, $5.00, $6.00, $7.00, $8.00, $8.50, $9.00, $9.95, $9.99, $10.00, $10.00, $10.00, $10.00, $13.00, $15.00, $15.00, $15.00, $17.00, $20.00, $22.00. Trials undercut standing rates and Community Tutors sit at the bottom, so a full hour lands around $10 to $30. The listings are public, and they'll have shifted by the time you read this.

The optional italki Plus subscription ($5.99 a month, or $4.99 billed annually) adds AI study tools. You don't need it to book lessons, and I wouldn't start with it.

Preply, and why I keep saying italki

Preply is the other big marketplace and I've been naming italki without explaining why, which isn't fair to either of them. The difference is the billing model, and it's structural rather than a matter of quality.

Preply runs on subscriptions. Per its own help centre, a subscription is tied to one tutor and renews automatically in 28-day cycles, with a minimum of 4 lessons per cycle. Lessons you don't schedule before the renewal date expire and don't carry over. You can adjust the plan, pause the renewal or cancel whenever you like.

italki sells credits. No subscription, no minimum, no renewal date, and switching tutors costs nothing because you were never committed to one.

So if you know you want 2 lessons a week with one person for the next 6 months, Preply's model matches that and the commitment is a feature: it's harder to quietly stop. If your weeks are irregular, or you want to try 4 tutors before picking one, an auto-renewing 28-day cycle with expiring lessons is the wrong shape and italki is the safer first move. That's the whole reason I default to it for a beginner, who by definition doesn't yet know which tutor suits them.

I'm not publishing hourly rates for Preply. Tutors set their own, the figures floating around come from third-party roundups rather than Preply, and a made-up number is worse than none.

Speaking for the first time is frightening in a way that has its own research behind it, and pretending otherwise doesn't help anyone. I've written about why speaking feels so much harder than studying separately.

All your free time is a commute: Pimsleur

A narrow segment, and a real one. If the only reliable half-hour in your day is behind a steering wheel, most of this page is useless to you, because it all needs a screen.

Pimsleur is audio-first and hands-free. And its method has genuine spaced-repetition ancestry, which is worth conceding plainly rather than treating as a marketing relic. Four named principles: anticipation, graduated interval recall, core vocabulary, organic learning. Graduated interval recall means learners "are reminded of new words at gradually increasing intervals," which is the idea a modern scheduler automates. Pimsleur was doing it on cassette decades before any of these apps existed.

Two consumer tiers, Premium for one language and All Access for all 50+ languages with up to 4 users, both with a 7-day free trial. Pimsleur's subscriptions page publishes no dollar figure, but its own blog says both cost "around $20 per month." That's the only first-party number I'll quote, and it's deliberately approximate. Anyone giving you $19.95 to the cent is reading a third-party blog.

The limit is that the content is fixed and slow by design. You get what the course gives you, in its order.

Stuck at intermediate: Anki

You finished a course app. You can read a menu, follow a slow podcast, and hold a conversation until it goes somewhere you weren't expecting. Then you stall, and doing more Duolingo doesn't move you.

The reason is that course apps have a finite syllabus and you've reached the end of what it can guess about you. What you need next is vocabulary chosen from the specific things you can't say, which no fixed course can know.

Anki ships no language content of its own. For a beginner that's a fatal flaw. For you it's the point: an empty scheduler you fill from your reading, your shows, and the words you reached for and missed yesterday.

Building the cards yourself is doing work, too. Slamecka and Graf named the generation effect in 1978: material you generate is remembered better than the same material read.

Here's the correction I'd most want an intermediate learner to have, because it's wrong in a hundred SEO posts. Anki's default scheduler is SM-2, Wozniak's 1987 SuperMemo algorithm. FSRS has been built into Anki since version 23.10 in October 2023 and it's excellent, but the manual calls it "an alternative to Anki's legacy SuperMemo 2 (SM-2) algorithm" and tells you to enable it at the bottom of the deck options page. Install Anki, never touch the settings, and you're running a formula from 1987.

Turn it on and you get FSRS-6: 21 parameters fitted to your own review history, with a desired retention target you set (the default is 90%). The mechanics are in the FSRS post.

Cost: free on Windows, macOS, Linux and Android, with free AnkiWeb sync. AnkiMobile on iOS is US$24.99 one time, which funds development. Regional prices differ.

Two things worth reading before you fill it. Which words to add first is a real question, and word frequency answers most of it (the frequency curve). And cards are only half of an intermediate plan, because at this level reading and listening volume does work that no deck does, which I've argued out in input versus cards. If you want the head-to-head on Anki specifically, that's Verbamor vs Anki.

Already paying for lessons: Anki, or Verbamor

This is the one segment where I'd put my own app on the shortlist, and it still isn't the automatic answer.

The problem is specific. You pay $20 for an hour, your tutor corrects six things, you understand all six in the room, and by the following Tuesday you can produce two of them. The lesson worked. The 6 days after it didn't.

The evidence for what fixes that is old and unusually clean. Karpicke and Roediger had students learn 40 Swahili-English pairs, then varied what happened after the first correct answer. Items that kept being tested came back at about 80% a week later. Items that were only restudied came back at about 36%.

Karpicke & Roediger 2008 · Swahili vocabulary, one week later

0%

recalled when items kept being tested after the first correct answer

0%

recalled when items were only restudied after the first correct answer

That last detail is the uncomfortable one. The students couldn't feel which method was working, which is why "I'll just look over my notes" survives contact with reality for so long.

Timing has numbers behind it too. Cepeda and colleagues, across more than 1,350 people, found the best-performing gap sat at roughly 20 to 40% of a 1-week test delay, falling to about 5 to 10% at a year. On a weekly lesson cycle that points at reviewing a day or two after you walk out, not the night before the next one. A week is at the very short end of what they measured, so read it as a direction.

Anki does this job for free, and for a tutored learner it does it well. The failure mode is arithmetic. A good card takes a couple of minutes to write, it gets written after the lesson when you're tired, and 2 skipped weeks build a backlog that ends the habit. If you're the person who types up 9 items every week for a year, you don't need me.

Verbamor is for the other person. It records the lesson, pulls out the words and phrases you got wrong, builds the cards from those, and schedules them with FSRS so each one returns near the day it would have gone. The material is your tutor's material.

Now the disclosure. Anki publishes a scheduler you can inspect: FSRS-6, 21 parameters, 90% default retention. Verbamor runs FSRS too, and there my evidence stops. I have no retention number of my own to show you. It hasn't been in enough hands for long enough to produce one that would mean anything, and a percentage invented for a comparison page is worth less than nothing. Everything measured on this page was measured by Karpicke, Cepeda and Slamecka, and it's evidence that retrieval and spacing work, not proof that my app runs them well.

Since I've published everyone else's prices, here's mine. Explorer is $19.99 a month. Native is $129.99 a year, about $10.83 a month, with the 7-day free trial on the annual plan. There's no free tier. It's iOS only. Both of those are real reasons to pick something else.

The longer version of this segment, with Quizlet and the homework question, is in language apps for tutored learners. And if you're weighing the free course app against lessons, Verbamor vs Duolingo covers that trade directly.

All 6, side by side

Why there's no test score on this page

The pages competing with this one tell you they tested everything. Some of them print per-app retention percentages from a 14-day trial with 50 cards, without naming a tester or a sample size.

I didn't run a study, so I'm not going to print one. What I can do is show my work: every factual claim here comes from a first-party page or a published paper, the prices are the ones the companies list, and where a number isn't verifiable I've said so rather than filling the gap.

That's why you'll notice some deliberate holes. I don't quote Duolingo's Energy starting amount, because Duolingo doesn't publish one. I don't give you Pimsleur to the cent. I don't tell you Babbel's interval ladder, because Babbel hasn't said. Every number in a competitor's table on those points was invented somewhere, and you can check that yourself in about 10 minutes.

Questions people actually ask

---

Sources

Every load-bearing claim Verbamor makes is traced to its paper on the research page. Prices and product facts on this page were checked against first-party sources on 11 August 2026.