Rosetta Stone vs Duolingo vs Babbel

Here is the state of this question. The page ranking first today lists a Rosetta Stone lifetime plan at $199.99 and an annual at $109.99. I opened rosettastone.com/buy on 21 September 2026 and there are two plans on it, priced $19.99 a month and $159 a year, and the word "lifetime" appears zero times.

Another page in the top three carries prices from January 2024 into a 2026 query. A third gives you "$8-$15/mo" ranges, which cannot be added up into what you'll be charged.

So this page starts with the prices, dated. Then it does something none of the three do, which is read the efficacy studies they cite instead of repeating the press release.

The 30-second answer

You came here for a pick, so you get one before the argument.

Spend nothing if you're just starting. Duolingo's free tier is a real free tier, not a trial: the core course, ad-supported and energy-gated. For a first few hundred words and a daily habit, it does the job for $0. A page that can't bring itself to recommend free is selling you something.

Pick Babbel if you want grammar explained. Lessons are built around short explanations in English, which suits people who want to know why a sentence works. Babbel publishes its review ladder, which almost nobody in this category does.

Pick Rosetta Stone if you want volume of speaking attempts and you're at peace with almost no English on screen. Its TruAccent speech tool is the oldest feature of its kind here, and one subscription covers all 25 languages, so it suits someone who will study two or three.

Pick none of them if you already pay a tutor. A course you're not following is money against a problem you don't have. That case gets its own section below, and it's the one I have a stake in.

How I checked this, and what I haven't done

I have not used these apps. There's no 60-hour trial behind this page and no team of testers, and you should be suspicious of comparison pages that claim one without showing you an artifact.

What I did instead is narrower and checkable. I read each company's own live pricing page on 21 September 2026 and quoted the figures on it. I downloaded the Babbel efficacy study PDF and read it, rather than quoting the press release about it. Every price below carries the date I read it, so when it drifts you'll know how stale this page is.

Two prices come with a caveat I'd rather state than hide. Babbel's plan table renders in the browser after the page loads, which is why automated checks miss it: the figures here were confirmed by opening my.babbel.com in a real browser on 5 September 2026 and are in our fact file with that date. And Duolingo publishes no billing term against its App Store prices, so I give a range rather than inventing a monthly figure.

I sell a language app. It's iOS-only, it costs money, and it does a different job from all three of these. I'll make that case once, near the end, after the useful part.

What they cost, read off their own pages today

Rosetta Stone, from rosettastone.com/buy on 21 September 2026. Monthly: $19.99, billed monthly. Yearly: $159, which the page displays as $13.25 a month with a "SAVE 33%" badge. Both cover all 25 languages. The page states a 14-day money-back guarantee, which is the only refund window any of the three puts in writing where I could find it.

The lifetime plan is the story. Comparison pages have quoted a Rosetta Stone lifetime at $199.99 for years, and it isn't on the pricing page now. Neither is the $109.99 annual. If you went looking for that deal on the strength of somebody's 2026 buying guide, you'd arrive at a page charging $159 and wonder what you did wrong.

Babbel, confirmed in a browser on 5 September 2026. One month $17.95. Three months $15.25 a month, billed $45.75. Six months $13.45 a month, billed $80.70. Twelve months $8.95 a month, billed $107.40. Lifetime $299.99. One subscription covers all 14 languages.

A warning that travels with every Babbel number: the site runs a countdown sale banner more or less continuously, and during that check a "Labor Day Sale" was offering lifetime at $199. The $17.95 monthly rate is the undiscounted reference Babbel computes its own "save 50%" badges against, so it's the honest one to plan from.

Duolingo has a genuinely free tier, and beyond that I'm going to be less precise than the pages you'll read elsewhere, on purpose. Duolingo's US App Store in-app purchases for Super run from $9.99 to $119.99, and the store attaches no billing term to any of them. Super carries a 14-day free trial on the US listing.

Which means the confident "$13.99 a month, $96 a year" you'll see quoted around is somebody's reconstruction. I can't confirm which price maps to which term from a first-party source, so I'm not going to print one. Duolingo Max pricing I can't confirm at all, so it isn't in the table.

Now the arithmetic that actually matters, because a per-month figure is designed to stop you doing it. A year of Rosetta Stone is $159. A year of Babbel, bought as the 12-month plan, is $107.40. Babbel's lifetime at $299.99 pays for itself against its own annual plan in under three years, and against Rosetta Stone's annual in under two.

The studies, and the one research team behind them

Every page ranking for this query cites the Babbel study. One of them puts it approvingly in its own section. Not one of them mentions who paid for it, so I downloaded it.

It's the Babbel Efficacy Study, Vesselinov and Grego, September 2016. Roumen Vesselinov is at Queens College, City University of New York, which is where the "CUNY study" framing comes from. Page 4 says this, verbatim:

Credit where it's due: they disclose it themselves, plainly, and that's better than plenty of published work manages. The problem is downstream, in the pages that quote the result and drop the sentence.

Two more things the study says that its fans don't. The famous 15 hours is not the headline figure. The executive summary reads: "Babbel users need on average 21 hours of study in a two-month period to cover the requirements for one college semester of Spanish." The 15 hours applies to "truly novice users with no knowledge of Spanish." So the number in circulation is the best case, quoted as the average.

And the test was WebCAPE, a written multiple-choice college placement exam that takes 20 to 25 minutes. The study measures where you'd be placed in a Spanish department, which is grammar and vocabulary recognition. I searched the full text for "oral" and "speaking" and got zero hits. The claim you'll read that 60% of participants improved their oral proficiency is not in this study; it belongs to separate later work.

The sample is worth knowing too. Participants were existing Babbel users invited by email, 391 in the initial random sample, 325 in the final study sample.

Then the part that reframes the whole category. That research team produced the efficacy evidence for most of this market. We counted 16 reports for 10 competing app companies between 2009 and 2025, including Rosetta Stone and Duolingo alongside Babbel, and the hub publishing them ranks the apps against each other in a table with no column for who paid. We wrote that up in one lab, fourteen apps, with the completion rates recalculated, because they range from 43.7% to 93.8% and the table shows sample size but never what the sample started as.

Duolingo sits bottom of that league table at 34 hours, on a 2012 study of a product that no longer exists. Its own report prints a median of 3.9 points an hour next to the mean of 8.1, and using the median doubles the hours. A 2021 paper in Foreign Language Annals, four of whose five authors work at Duolingo, says the reports were "published on company websites as white papers, not peer-reviewed journal articles" and that comparing effectiveness across the products is hard.

So: three efficacy claims, one research team, no independent head-to-head. That's the honest summary, and it applies to all three apps equally.

What a year actually buys you

Here's the question a subscription page won't answer: how many hours of material is behind the paywall, and what happens at the end of them?

Babbel is the only one of the three to put a number on its own content, and it's a big one. Babbel says it has "60,000 lessons across 14 languages" and that all 100,000 hours of its content is voiced by native speakers. Lessons run about 10 to 15 minutes by its own description, though it gives 6, 10-15 and 15 minutes on different pages, so treat that as approximate.

Duolingo publishes a ceiling instead of an hour count, and it moved recently. Since 22 April 2026 it says nine of its courses, Spanish included, teach through CEFR B2, a Duolingo Score of 129 on a scale that tops out at 160. Any page still telling you Duolingo caps out at A2 is repeating something that stopped being true in April. That's a claim about what's in the course, not about what learners come out able to do.

For a sense of the real distance: the Foreign Service Institute puts Spanish and French at 30 weeks and 552 to 690 class hours to reach a working professional level, on a week of 23 class hours plus 17 hours of self-study. We did that arithmetic properly in the FSI hours post, including why almost every blog still quotes the retired figure. Against that scale, 21 hours of anything buys a college placement bump, which is exactly what the study measured and no more.

The thing none of the three tell you is where finishing the course leaves you. It's the point the maintenance problem starts, and maintenance is a different activity from instruction.

Whether any of them gets you speaking

All three have a microphone. Rosetta Stone's TruAccent is the best known, Babbel has speech recognition in its lessons with full coverage only in Chrome, and Duolingo has speaking exercises from early on.

They share a design property worth knowing about before you pay for one. In each case the sentence is on the screen before you say it. You're reading aloud and being scored on how it sounded.

That's genuinely useful for pronunciation, and it's a different exercise from producing a word from memory with nothing in front of you. Kang, Gollan and Pashler tested exactly this in 2013: learners who tried to produce a Hebrew word from memory before hearing it beat learners who repeated after a model, on a spoken test two days later, with no loss of pronunciation quality. Their production test was naming single pictured objects, so don't stretch it to conversational speed, but the direction is clear.

Which is why the speaking row in the table below says "yes" for all three and then says what kind.

The comparison, side by side

That Babbel review row deserves a sentence of its own, because it cuts both ways. Babbel is the only one of the three that publishes its interval ladder, which is more honesty than the category usually offers. And the ladder has six steps and then stops, the same schedule for every learner, and you cannot put your own words into it.

What happens when the course runs out

Every scenario above ends inside one of these three apps, which is where the other comparison pages stop too. So here's the part they leave out.

A course teaches you a curriculum. What it can't do is catch the specific word you reached for on Tuesday and couldn't find. That word came from a conversation, or a show, or a tutor correcting you, and no fixed course knows it happened. Babbel says this about itself, in its own help centre: review vocabulary is "predefined based on the lessons you complete. It is not possible to manually add random words or phrases to your review."

For most people the answer is a notebook, and I mean that seriously rather than as a setup. A notebook of words you personally got stuck on, revisited, beats a fourth app you open twice. If you'll keep one, keep one.

If you want the words scheduled rather than piled up, Anki is free on desktop and Android, costs $24.99 once on iOS, and has one of the best review engines ever shipped. Its card library is empty on purpose, so you type every card yourself. People who enjoy that get the best deal in this entire article.

Verbamor, which I built, does one narrow thing: it records a lesson or takes a phrase you typed, pulls out the words and phrases you actually used or fumbled, and turns them into fill-in-the-blank cards scheduled by FSRS, with generated audio and an image on each. Five languages: Spanish, French, Italian, Portuguese and Russian. $19.99 a month or $129.99 a year, 7-day trial on the annual plan, iOS only, no free tier. Both are US prices, read off the app's own plan definitions on 21 September 2026; the App Store sets a different number in other countries.

It teaches you nothing. There's no curriculum in it, and on day one it's an empty room. Against these three apps for a beginner it loses outright, and I'd rather say so here than have you find out after paying. It earns its place only once you have a source of new words that a course can't see.

Questions people actually ask

---

Sources

Every load-bearing claim Verbamor makes is traced to its paper on the research page.