Best Anki Settings for Language Learning (FSRS, 2026)

The best Anki settings for language learning in 2026 are short: turn on FSRS, set desired retention to 0.85 to 0.90, cap new cards at 15 to 25 per day, leave learning steps at 1m 10m, and set maximum reviews per day to 9999 so the algorithm can actually show you everything that is due. Do that and you have a deck that schedules itself around your memory instead of fighting it.

Everything below explains why each of those numbers is what it is, and the handful of settings people get wrong that turn Anki into a daily chore they quit.

First, turn on FSRS

FSRS (the Free Spaced Repetition Scheduler) is built into modern Anki and is the single most important setting on this page. The old SM-2 scheduler used fixed multipliers: get a card right and its interval roughly doubles, regardless of how hard that specific card is for you. FSRS instead builds a mathematical model of your personal forgetting curve from your actual review history, then schedules each card for the day you are about to forget it.

For language learners that difference is large, because vocabulary is wildly uneven. A transparent cognate and an irregular verb do not deserve the same interval, and FSRS stops treating them as if they do.

To enable it: open any deck’s options, go to the FSRS section, and toggle it on. It applies across your whole collection, not just one deck.

Desired retention: aim for 0.85 to 0.90

This is the one setting most worth understanding. Desired retention tells FSRS what fraction of cards you want to remember when they come due. The default is 0.90 (90%), and the practical range is roughly 0.80 to 0.95.

It is a dial between effort and forgetting:

Setting	What you get
0.80	Fewer reviews, but you forget more words before they stick
0.85 to 0.90	The language-learner sweet spot: solid recall, sane workload
0.95+	Near-total recall, but review count climbs steeply
0.97+	Workload becomes overwhelming for diminishing returns

For vocabulary, where approximate recall is usually fine and you will see common words again in real input anyway, 0.85 to 0.90 is the range to live in. Below 0.80 you start forgetting words before they consolidate; above 0.95 you are grinding for tiny gains. If your daily reviews feel crushing, drop retention by 0.02 to 0.03 rather than abandoning the deck.

New cards per day: 15 to 25, every day

The number that wrecks more learners than any other is new cards per day, because enthusiasm makes people set it too high. Every new card you add today becomes 8 to 10 reviews spread across the coming weeks. Add 50 a day for a week and you build a review debt that collapses on you.

Set new cards to 15 to 25 per day and keep it there. Consistency beats volume: 15 cards every single day is far better than 50 cards for three days and then a week off. FSRS works equally well whether you learn 5 or 50 new cards a day, so the limit is about protecting your future self’s review load, not the algorithm.

If you want the deeper version of why showing up daily beats heroic weekend sessions, see our guide on how to stay consistent learning a language.

The smaller settings that still matter

Learning steps: 1m 10m. With FSRS doing the long-term scheduling, learning steps only handle a card’s first few minutes. Short steps like 1m 10m let a new card graduate the same day and hand it off to FSRS. Long ladders of learning steps (the old 1m 10m 1h 1d 3d style) just fight the scheduler now.

Maximum reviews per day: 9999. This one is counterintuitive. The max-reviews cap silently hides cards that are genuinely due, so a low value means you are skipping reviews FSRS scheduled for a reason, which quietly wrecks your retention. Set it high (9999) and control your workload through the new-card limit instead, which is the lever that actually shapes future review volume.

Use one preset per group of similar material. Presets hold FSRS parameters. Cards that behave similarly should share a preset so FSRS has enough data to model them well. All your Japanese vocabulary can share one preset; a Japanese deck and a geography deck should not, because they are learned differently. Do not make a separate preset for every deck, and do not lump unrelated subjects together.

Setting	Recommended value
Scheduler	FSRS (enabled)
Desired retention	0.85 to 0.90
New cards/day	15 to 25
Learning steps	1m 10m
Maximum reviews/day	9999
Presets	One per group of similar material

Optimize your parameters (and keep doing it)

FSRS ships with sensible default parameters, but it gets noticeably better once it learns from your reviews. In Anki 24.06.3 and newer, the Optimize button works with any number of reviews, so you can run it early and often.

In the FSRS settings, click Optimize. Anki analyzes your review history and tunes the model’s weights to your memory. Re-run it about once a month, or each time your review count doubles (after 100 reviews, then 200, then 400, and so on). Early on, optimizing more often helps the model settle; later, monthly is plenty.

Settings won’t fix bad cards

The fastest way to hate Anki is to fill it with cards that are vague, overloaded, or copied from someone else’s deck. No scheduler can rescue a card you do not understand. A few habits that matter more than any toggle:

• Keep cards atomic. One fact per card. A sentence with three new words should be three cards, not one.
• Test recall, not recognition. Prefer producing the word from a meaning or an example sentence over just reading it.
• Add context. A bare word-to-word pair is brittle. An example sentence gives your memory a hook.
• Delete cards you keep failing. A card you have lapsed five times is usually a bad card. Rewrite or bury it; do not let it punish you daily.

This is exactly where spaced repetition fits into the bigger picture. It is one of a handful of methods that genuinely move the needle, which we cover in the best language-learning methods that actually work.

Is a dedicated app sometimes better than configuring Anki?

For kanji specifically, the build-it-yourself freedom of Anki competes with a tuned, batteries-included SRS like WaniKani. If you are weighing the raw flexibility (and zero cost) of Anki against a system that makes every decision for you, our breakdown of WaniKani vs Anki for kanji walks through who each one suits.

Common questions

What is the best desired retention for Anki language decks? Stay between 0.85 and 0.90. It keeps recall strong without ballooning your daily reviews. Lower it slightly if reviews feel unsustainable; raise it only if you truly need near-perfect recall and can absorb the extra workload.

How many new cards a day should I do in Anki? 15 to 25, done every day, is the sustainable range for most learners. The exact number matters less than doing it consistently, since each new card generates many future reviews.

Is FSRS better than the default Anki scheduler? Yes, for almost everyone. FSRS models your personal forgetting curve instead of using fixed interval multipliers, so it schedules each card more accurately and usually cuts your review count for the same retention.

Do I need an add-on to use FSRS? No. FSRS is built into current versions of Anki. Just enable it in the deck options and click Optimize once you have some review history.

How often should I optimize FSRS parameters? About once a month, or whenever your total reviews double. Optimizing more often early on helps the model learn your memory faster; after that, monthly is enough.

Why is my Anki review count so high? Almost always too many new cards per day, a desired retention set too high, or a max-reviews cap that was hiding cards that are now overdue. Lower new cards first, then nudge retention down by 0.02 to 0.03 if needed.

The takeaway

Good Anki settings are not exotic: FSRS on, retention around 0.85 to 0.90, 15 to 25 new cards a day, max reviews wide open, and a monthly Optimize click. Get those right, write atomic cards you actually understand, and show up daily. The algorithm will quietly do the rest while you go enjoy real input in your target language.