Russian flashcards, and the 33 letters nobody teaches first

Здравствуйте is 12 letters and it means hello. If you can't read it, a flashcard with that word on the front tests one thing: whether you can match a squiggle to another squiggle you saw yesterday.

That's the state most people are in when they search for Russian flashcards, and it's the exact thing the three pages ranking for it leave out.

Flashcardo gives you the 1,000 most common Russian words in 20 blocks of 50, credited to Pinhok Language's Russian Frequency Dictionary, plus 16 topic decks and printable PDFs in 8 formats. Speechling gives you a free sentence tool with male and female native voices across 11 graded modules. Quizlet gives you an anonymous user set of terms and definitions with no prose attached at all.

All three read on 21 September 2026. Zero mentions of the alphabet across the lot.

Can I use Russian flashcards before I can read Cyrillic?

Not really, and the failure is quiet enough that people blame themselves for it.

Here's what actually happens. You meet ресторан in a deck, you can't decode it, so your brain stores the overall shape. The card comes back, the shape matches, you tap Good. Ten cards in, three of them start with the same shape and the whole thing collapses. Then you stand in front of a real sign and the word doesn't arrive, because you learned pictures of words.

Cyrillic is 33 letters. It's phonetic, spelling is far more regular than English, and the shapes go in over a few days. The reason it feels hostile is a smaller and more specific problem: 7 letters look like Latin letters and lie about it. В is v. Е is ye. Н is n. Р is a rolled r. С is s. У is oo. Х is a throaty kh.

Those 7 are why a beginner reads РЕСТОРАН as "pectopah" with total confidence and walks past dinner. A letter you half-know slows you down. A letter you confidently know wrong produces a word you say out loud in a shop.

The useful ordering trick is to front-load the 5 letters that look Latin and behave, because А, К, М, О and Т already spell МАМА, КОТ and ТАМ. Three real Russian words off 5 letters, on day one. The full 33 with a hook for each one is in the Cyrillic alphabet guide, and it's free, and you don't need an app for it.

What you do need is a way to know when you're done. "Feeling comfortable with the alphabet" is exactly the judgment your brain is worst at making, because you're judging it with the letters in front of you. Koriat and Bjork called that an illusion of competence in 2005.

A scheduler can answer it instead. In our own alphabet deck the threshold is 4 days of stability on every letter, in both directions, before the app suggests moving on to words. Stability is an FSRS term for how many days it takes your predicted recall of an item to fall to 90%, so 4 days across all 33 means the model thinks you'd still have the whole alphabet at the end of a long weekend. Whether you use our deck or index cards, that's the shape of the question to ask: would I still have these on Friday.

A pile of cards is not a schedule

This is the argument all three competitors leave on the table, and it decides whether any of those 1,000 words are still in your head in March.

Two findings do all the work.

Retrieval beats restudying. Karpicke and Roediger taught students 40 Swahili-English word pairs and tested them a week later. In the conditions that kept retrieving a pair after it was first recalled, students got about 80% of the list. In the conditions where pairs were dropped from testing and only restudied, 36% and 33%.

Spacing beats massing. Cepeda and colleagues reviewed 317 experiments on distributed practice in 2006. The same total study time, chopped across days, buys more memory than the same time in one sitting.

Karpicke & Roediger 2008 · one week later

0

% recalled when pairs kept being tested after the first correct answer

0

% recalled when pairs were repeatedly restudied instead, and dropped from testing

Now look at what's on offer. Speechling's tool has no scheduler at all, so nothing decides when a word comes back and nothing records that you forgot it. Quizlet's help centre documents no spaced repetition scheduler either: what ships is Progress, a mastery tracker sorting terms into Not Studied, Still Learning and Mastered with no dates attached, and the modes that do drill you are gated. Flashcardo has an online spaced repetition tool, the best of the three, and it asks you to register before you've seen anything work.

So a 1,000-word catalog with no schedule behind it is 1,000 words you'll meet once and lose. The catalog was never the hard part.

The scheduler worth knowing by name is FSRS, built by Jarrett Ye and the Open Spaced Repetition community. It tracks difficulty, stability and retrievability per card, with 21 parameters fitted to your own review history. One caveat I'd rather state than skip: the widely repeated claim that it needs 20 to 30% fewer reviews than the old SM-2 formula comes from FSRS's own documentation rather than an independent trial.

Frequency order manufactures the confusion it ignores

Flashcardo's sequencing is the most thought-out on this search and I want to credit it properly. Ordering by corpus frequency answers "what do I learn first" with data rather than with somebody's hunch that colours are a beginner topic. It beats the topic decks sitting directly above it on the same page.

It's also incomplete, in a way that costs you specifically.

Words from one category are harder to learn when they arrive together. The learner builds one fuzzy node and then guesses inside it, which is why два and три swap places in your head for a month. Tinkham found it in 1993 and again in 1997, Waring replicated it in 1997, Erten and Tekin got the same result with young learners in 2008, and Nation wrote the guideline in 2000: teach one, let it settle, then the next.

A frequency list walks straight into this. Numbers are frequent, so the numbers cluster. Common verbs of motion cluster. Pronouns land in a run. Nothing in the corpus knows which pairs confuse in a head, because frequency is a property of the language and interference is a property of you.

I'd rather not oversell it. Higa (1963) found no penalty for antonym pairs, and Nakata and Suzuki (2019) found spacing itself shrinks the interference, which means a decent scheduler buys back some of what a bad order costs. The fix is cheap rather than dramatic: spread the members of a confusable set apart, and put core verbs and function words first.

For our own starter decks we tag the sets by hand and hold their members apart, with a test in the build that fails if any two members drift closer than the minimum gap. A catalog of 1,000 frequency-ranked words can't do that, because nobody has tagged which of those 1,000 confuse with which.

What goes on one Russian card

Every page on this search describes its cards. Here's one instead, so you can argue with the actual thing.

That third row is what separates Russian from the other languages these template sites cover.

You learn человек, person. Good. Then somebody says много людей and the word is unrecognisable, because Russian marks grammatical role on the noun itself and the plural genitive of человек is suppletive. Knowing the dictionary form didn't give you that. A card for the form, labelled as the genitive plural, is a separate thing you retrieve separately.

Frequency lists give you headwords. The forms have to come from sentences you actually met, which is a structural reason a catalog can't supply them.

Two more design choices have sources, and one has a caveat the marketing pages skip.

The audio. Bassetti and Atkinson (2015) found the written form still bends pronunciation in learners with years of instruction, which argues for hearing a word every time you meet it, doubly so when the script is new. Ours is synthesised speech generated per card with a native-accent instruction. Speechling's is recorded native speakers in two voices, and on that axis Speechling wins outright.

The image, with a caveat. The picture-on-a-card claim gets overstated everywhere, including by people selling cards. Carpenter and Olson (2012) tested exactly this on foreign vocabulary and found pictures did not initially beat native-language translations, and made learners more overconfident about words they'd seen with a picture. The benefit appeared only once that overconfidence was corrected by retrieval practice or an explicit warning. So an image helps when it isn't allowed to make you feel like you already know the word.

What all of these cost, including mine

This is where the search fails buyers worst, so here is everything I could confirm on a first-party page on 21 September 2026, plus the things I couldn't.

Two notes, because an honest price table has to include what's missing from it.

Quizlet's own help centre confirms the gating above in its own words, and confirms that the class bypass is real. What Quizlet won't state on a page I can fetch is a number. Third-party sites print figures confidently and I won't repeat one I can't trace, because a wrong price would discredit every other row here. Open the app and read it yourself before you subscribe.

A second thing about Quizlet, if you use an AI assistant to research purchases. Its pages return HTTP 403 to anything that isn't a browser, which I confirmed again today. So a model summarising this search for you is working from what other people wrote about Quizlet, at second hand, with no way to check.

And we're the expensive one on that list by a distance, so let me address it rather than hope you don't notice.

You're not buying a bigger catalog from us. What the money buys is the card-making, which is the expensive part of every flashcard habit. You record or paste what came up in your lesson, and the words and phrases you actually reached for come back as cloze cards with audio and an image, scheduled by FSRS. Gabriel Wyner, who has probably made more language cards by hand than anyone reading this, puts card creation at 2 to 3 hours a week to sustain a 25-a-day habit. That's the line item you're removing.

Honest limits, since they matter more than the features. It's iOS only, no Android and no web. It teaches you nothing new, it only holds what your lesson taught you, so it won't carry a total beginner through their first thousand words. And it's 5 languages, Russian among them, rather than a general flashcard tool.

What flashcards cannot fix

All three of these pages sell cards as the whole answer. Cards fix retention. They don't fix production, and the gap between those two is where most Russian learners live.

You know the word. It's in the deck, you passed it Tuesday, and mid-sentence it doesn't arrive. A catalog can't design around that, because it has no idea which words you dropped.

The words worth carding are the ones you personally fumbled while speaking, which is a different harvesting strategy from downloading a top-1,000 list. It's the whole reason I built what I built: you record the lesson, and the words you reached for and missed become the cards. A word you already own but stumbled over gets its due date pushed into the past, so it surfaces at the top of the next queue instead of sitting three weeks out on a schedule that no longer matches reality.

Your review load is set by your new-card rate, too. Anki's own manual puts the ratio plainly: "If you are consistently learning 20 new cards a day, you can expect your daily reviews to be roughly about 200 cards/day." At 5 new cards a day that survives. At 20 it's a part-time job. More in how many flashcards a day.

And run input alongside the deck. Nation (2006) found unassisted comprehension of spoken text at 98% coverage takes 6,000 to 7,000 word families, and he has a line about 95% coverage I think about a lot: roughly 7 unknown words per minute of speech at a normal 150 words a minute. No deck closes that alone. The split is argued out in comprehensible input vs. flashcards.

Go use Anki or Flashcardo if this is you

Two groups should stop reading and go elsewhere, and I mean that literally rather than as a rhetorical move.

If you enjoy making cards, use Anki. Free on Windows, macOS, Linux and Android, syncing through AnkiWeb is free, and iOS is US$24.99 once rather than a subscription. It ships no Russian of its own, which is the catch and the point: you're buying an engine and supplying the words. Turn FSRS on at the bottom of the deck options page, because Anki still defaults to SM-2 from 1987 and plenty of articles wrongly say otherwise. For someone who likes the craft, Anki is the right answer to this entire page and nothing paid beats it.

If you're starting from zero with no tutor and no lessons, use Flashcardo. It's free, the top-1,000 list is a defensible starting order, and the printable PDFs in 8 formats are genuinely useful at the alphabet stage. We hold on to words you've already met, which makes us the wrong tool for a learner who hasn't met any yet.

The case for paying for something like ours starts where you're having real conversations in Russian, losing words out of them, and finding that no catalog contains the specific words you keep losing.

Questions people actually ask

---

Sources

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