Anki FSRS for Language Learners: Why the 2026 Scheduler Can Cut Reviews 20–30% vs. SM-2
See what FSRS’s 20–30% fewer-review estimate really means, then switch from SM-2 safely and tune Anki workload for language learning.
Current 2026 FSRS simulations estimate roughly 20–30% fewer reviews than SM-2 at the same retention, but your real savings vary—and fewer reviews are not automatically the same percentage less study time.
If your Anki deck has gone from helpful notebook to demanding landlord, FSRS is worth switching to—but not because it contains a magic “30% off” button. The gain comes from scheduling reviews around a better estimate of forgetting and letting you choose the retention/workload tradeoff explicitly.
For most language learners, the useful FSRS setup is simpler than the algorithm sounds:
- Enable FSRS.
- Optimize the parameters from your own review history.
- Start around Anki’s default 90% desired retention instead of chasing the highest number.
- Use Again when you actually forgot. Hard still means you recalled the answer.
- Do not mass-reschedule your entire old SM-2 collection on day one unless you understand the due-card spike it can create.
- Judge the result from monthly retention and workload—not one ugly Tuesday.
First: what does “20–30% fewer reviews” actually mean?
The strongest current numerical comparison comes from the FSRS project’s 2026 ABC of FSRS documentation. It reports that simulations updated on February 12, 2026 produced about 20–30% fewer reviews than SM-2 at the same retention level.
Three words in that sentence do a lot of work: simulations, reviews, and same retention.
- Simulations: this is a benchmark, not a promise that your personal deck will drop by exactly 27% next month.
- Reviews: it is a count of repetitions, not guaranteed clock-time savings. A five-second vocabulary recognition card and a 25-second listening-production card are both “one review.”
- Same retention: the comparison is meaningful only if you are trying to remember roughly the same proportion of material.
Anki’s own Deck Options manual uses safer wording: FSRS can help you remember more material in the same amount of time. Its official scheduler FAQ also says FSRS needs fewer reviews than the legacy algorithm to reach the same retention, without promising a universal percentage.
So the headline is useful—but treat it like a fuel-economy estimate, not a coupon.
Why FSRS can beat SM-2 without “trying harder”
SM-2 is Anki’s legacy scheduler. FSRS is a newer scheduler that estimates how likely you are to remember a card and schedules around that estimated memory state. You do not need to understand the equations to benefit from the practical difference.
For a language learner, imagine two Spanish cards:
- perro → dog, which you have answered instantly for months;
- a pesar de que → although/even though, which you still hesitate over in sentences.
A scheduler should not keep charging both cards the same “review rent” just because they entered the deck around the same time. FSRS uses your history to estimate memory more flexibly, then schedules toward a retention target you choose.
That target is where the article stops being about clever algorithms and becomes about your daily life.
The Workload Dial: desired retention is not a score to maximize
In Anki’s current FSRS settings, desired retention is the central workload control. The default is 90%. Anki’s manual says higher retention gives shorter intervals and more reviews, and warns that workload climbs sharply as you move higher—especially toward the upper 90s.
Think of desired retention as a dial, not an exam score:
- Turn it up: you should forget fewer cards, but Anki needs to show them more often.
- Turn it down: you get fewer scheduled reviews, but more cards will be forgotten when they return.
There is no medal for setting 97%. If 97% makes Anki consume the time you used to spend reading, speaking, watching shows, or actually using the language, you have raised your own rent.
Which situation describes you?
“My reviews are manageable and my monthly retention is close to target.”
Leave the dial alone. A stable system is not improved by constant fiddling. Keep reviewing honestly and optimize FSRS parameters periodically.
“Anki is crowding out real language use.”
First reduce the number of new cards you introduce. If the queue is still excessive, inspect your actual retention/workload data before making a modest retention adjustment. Lower retention means accepting more forgetting; it is a trade, not free time.
“I’m forgetting too much.”
Before increasing desired retention, check two things: are you grading forgotten cards as Again, and are your cards actually answerable? Ambiguous cards can look like a scheduling problem when the prompt is the real problem. If those are clean, a modestly higher retention target may make sense.
“I just switched from SM-2 and have no stable data yet.”
Start conservatively. The 90% default is designed as a balance between retention and workload. Give FSRS time to work with your history before tuning the dial because one dramatic day is terrible evidence.
How to switch from SM-2 to FSRS without detonating your queue
Anki’s current manual gives a straightforward migration path. For a typical language learner:
That final checkbox matters. The manual notes that rescheduling when first moving from SM-2 can make a large number of cards immediately due, depending on your desired retention. FSRS did not suddenly decide you deserve punishment. You asked it to rewrite the calendar all at once.
The grading rule that can wreck FSRS: forgotten means Again
This is the easiest rule in the article and probably the most important:
If you did not recall the answer, press Again.
Under FSRS, Hard is still a successful recall—hesitant, effortful, ugly, but successful. Anki explicitly warns that using Hard when you actually forgot can produce intervals that are unreasonably long.
Language learners are especially tempted to blur this because partial familiarity feels seductive.
Four quick grading situations
You see the German noun die Entscheidung. You know it means “decision,” but you cannot recall whether the article is die or der.
If the card specifically tests gender + noun, you failed the requested information. Press Again. “I knew the English meaning” is not successful recall of that card.
You see a Japanese kanji reading. The answer comes after five uncomfortable seconds, but you produce it before revealing the back.
That can legitimately be Hard: you recalled it, but with substantial effort.
You reveal the answer and think, “Oh! Of course. I totally knew that.”
Recognition after seeing the answer is not recall. Press Again. Hard is not “I forgot, but romantically.”
Your Spanish sentence card allows three reasonable translations, but the back accepts only one.
This is partly a card-design problem. Grade the current attempt honestly, then rewrite the card so the next review tests something unambiguous.
FSRS cannot rescue bad language cards
A better scheduler decides when to test you. It cannot decide whether your prompt is worth testing.
Before blaming FSRS for a stubborn card, check for these language-learning traps:
- Too many answers: “Translate take” is not a card; it is an ambush.
- Recognition pretending to be production: seeing the target word and remembering its meaning is easier than producing the target word from meaning/context.
- Overloaded cards: one card asks for spelling, pronunciation, gender, collocation, translation and a sample sentence.
- Slow media cards: a listening card may take 20–30 seconds, so fewer reviews will not translate one-for-one into fewer minutes.
- Leeches: if the same item keeps failing, rewrite, split, add context, or suspend it instead of paying review rent forever.
The scheduler is a traffic controller, not a language teacher.
The 10-card grading audit
On your next session, pay special attention to ten difficult cards.
- Before revealing the back, ask: Did I retrieve the information the card actually asked for?
- If no, use Again.
- If yes but it took serious effort, Hard may fit.
- If the prompt had multiple reasonable answers, mark it for rewrite.
- If you keep failing the same item despite a clear prompt, consider changing the card or suspending it.
This tiny audit often fixes more than another hour of reading about FSRS mathematics, because good scheduling depends on honest review data.
How often should you optimize FSRS?
You do not need to press Optimize after every heroic study session. Anki’s manual says once a month is sufficient. The optimizer learns from your review history and produces parameters that better fit your memory and material.
It also warns against manually changing or copying FSRS parameters. Someone else’s optimized vocabulary deck is not your brain, your card design, or your review behavior.
If your decks differ dramatically—for example, five-second word-recognition cards versus slow listening-production cards—separate presets can make sense because the material behaves differently. Do not create seventeen presets because Chapter 8 “felt kind of hard.”
How to prove FSRS is helping your own language deck
The 20–30% figure is a useful benchmark. Your own data is the decision.
After you have used FSRS consistently for a while, check three things monthly:
- True retention: is your actual pass rate reasonably close to desired retention?
- Workload: how many reviews or minutes are you spending per day?
- Language time: is Anki leaving enough room for reading, listening, conversation and output?
Anki’s current manual includes a simulator that can estimate workload in reviews per day or minutes per day at different retention settings. That is far more useful than changing retention because Reddit, a YouTuber, or your emotionally exhausted 11:40 p.m. self said 92% sounded optimal.
Only change the Workload Dial after you know what problem you are fixing.
Language learners: the fastest way to reduce reviews is sometimes fewer new cards
FSRS schedules the cards you give it. It does not stop you from adding 70 new words from every episode of a show.
If your review count is exploding, ask a brutal question: Would I care about this word if it had not appeared on screen five minutes ago?
Keep high-value items: recurring words, useful chunks, phrases you want to say, grammar patterns you keep missing, and vocabulary tied to meaningful scenes. Be suspicious of one-off nouns that merely happened to exist near a dramatic explosion.
This is where spaced repetition fits inside media-based language learning: the media gives you context and motivation; Anki protects selected items from disappearing. The flashcard queue should not become a transcript landfill.
Where FunFluen can fit before Anki
If you learn from supported video pages, FunFluen can help you collect contextual language items and, where your account/data supports it, export saved items for external review. That is an input-quality job; Anki still owns FSRS scheduling and intervals.
If that workflow is useful to you, review the FunFluen browser extension before installing. Export options depend on available saved data and account/tier, and there is no claim here that FunFluen runs or syncs FSRS.
FSRS FAQ for language learners
Should I set desired retention above 90%?
Only if the extra retention is worth the extra workload to you. Anki’s default is 90%, and its manual warns that reviews increase sharply as retention rises. Higher is not automatically better if it crowds out actual language use.
Should I enable Reschedule Cards on Change when switching from SM-2?
Usually not as your first move. Leaving it off means future reviews use FSRS without immediately rewriting every existing due date. If you intentionally enable it, back up first and expect that many cards may become due at once.
How often should I optimize FSRS parameters?
Current Anki guidance says monthly optimization is sufficient. You do not need to optimize constantly.
The one-month FSRS challenge
Switch cleanly. Keep the 90% default unless you have a reason not to. Grade forgotten cards Again. Fix ambiguous prompts. Add fewer junk cards. Optimize when appropriate.
Then leave the Workload Dial alone for a month and measure what actually happens.
If FSRS gives you 20–30% fewer reviews, wonderful. If your reduction is smaller, but the queue is steadier and your retention is where you want it, that can still be a win. And if Anki is still eating your whole evening, the next question may not be “Which scheduler?” It may be “Why am I feeding this thing so many cards?”
That is the real upgrade: not worshipping a smarter algorithm, but getting your language study back from the queue.