FSRS vs SM-2 (2026): Which Anki Algorithm Wins? Benchmarks Compared
The old standard vs the modern scheduler, by the numbers, and when each one still makes sense
TL;DR
Quick verdict: <strong>FSRS</strong> reaches the same retention as <strong>SM-2</strong> with roughly 20–30% fewer reviews because it models three memory variables instead of one rigid formula. SM-2 is simpler and battle-tested; FSRS is data-driven and adaptive. For most learners in 2026, FSRS is the better default, and Flica ships it on by default.
SM-2 is the 1980s SuperMemo algorithm that powered spaced repetition for decades, including classic Anki. FSRS (Free Spaced Repetition Scheduler) is the modern open-source successor, trained on hundreds of millions of real reviews. The practical question is simple: for the same retention, which one makes you do fewer reviews?
This guide explains how each algorithm decides when to show a card, why FSRS is consistently more efficient in published benchmarks, the cases where SM-2 is still fine, and exactly how to switch. No math degree required.
How SM-2 Works
SM-2 tracks an ease factor per card and multiplies the interval by it after each successful review. Get a card right, the interval grows by the ease factor; get it wrong, the card resets to the start. It is elegant and predictable, but it assumes every card behaves the same way and adjusts only through one coarse number.
- One variable, an ease factor that scales intervals
- Reset on lapse, a single miss sends a card back to the beginning
- Fixed rules, the same formula for every card and every learner
- Transparent, easy to understand and reason about
SM-2's weakness is rigidity: it can't tell that one card is intrinsically hard for you and another is easy, it treats them with the same formula.
How FSRS Works: The DSR Model
FSRS models memory with three values per card, Difficulty (how hard this card is for you), Stability (how long the memory lasts before it decays), and Retrievability (the probability you can recall it right now). It schedules each review for the moment your recall probability drops to your target (e.g. 90%), then updates the model from your actual answer.
- Difficulty (D), per-card, learned from your performance
- Stability (S), how slowly the memory decays
- Retrievability (R), current recall probability, drives timing
- Data-trained, parameters fit hundreds of millions of reviews
Because FSRS predicts recall per card instead of applying one fixed rule, it schedules each card at a smarter time, fewer wasted early reviews, fewer forgotten cards.
The Efficiency Difference
In Anki's own benchmarks and independent analyses, FSRS achieves the same retention target as SM-2 with roughly 20–30% fewer reviews. Over a semester of a few thousand cards that is hundreds of reviews, and hours, saved, with equal or better retention.
| Dimension | SM-2 | FSRS |
|---|---|---|
| Variables modeled | 1 (ease factor) | 3 (D, S, R) |
| Per-card adaptivity | Low | High |
| Reviews for same retention | Baseline | ~20–30% fewer |
| Configurable retention target | No (implicit) | Yes (e.g. 90%) |
| Data-driven parameters | No | Yes (trained) |
| Transparency | Very high | Moderate |
When SM-2 Is Still Fine
SM-2 is not broken. For small decks, short timelines, or learners who value a perfectly predictable schedule, the efficiency gap barely matters. It also has decades of real-world use and is trivial to reason about.
- Small or short-lived decks, the savings are negligible
- You want predictability, SM-2's intervals are easy to forecast
- Legacy setups, existing SM-2 history that you don't want to disturb
If you have years of SM-2 history in Anki, FSRS can use that history when you enable it, you don't start from zero.
How to Switch to FSRS in Anki
FSRS is built into modern Anki. You enable it in deck options, optionally optimize parameters on your review history, and set a target retention. The switch is reversible and uses your existing cards.
- Update Anki, recent versions include FSRS natively
- Deck Options → enable FSRS, toggle it on
- Optimize, fit parameters to your own review history
- Set target retention, 90% is a sensible default
Prefer not to manage any of this? Apps like Flica ship with FSRS enabled and tuned by default, there is nothing to configure.
FSRS Without the Setup: Flica
Flica uses FSRS by default, no toggles, no optimization step. It also generates cards with AI from a YouTube link, PDF, or text, so the modern algorithm and modern card creation come together.
- FSRS on by default, optimal scheduling from the first review
- AI card generation, YouTube, PDF, or text into cards
- Free on iOS and Android, core features at no cost
- Full offline review, study anywhere
Flica is on the App Store (https://apps.apple.com/app/flica) and Google Play (https://play.google.com/store/apps/details?id=app.flica).
FAQ
Is FSRS better than SM-2?
For most learners, yes. FSRS reaches the same retention as SM-2 with roughly 20–30% fewer reviews because it models difficulty, stability, and retrievability per card instead of using one fixed ease factor. SM-2 remains fine for small decks or when you want a perfectly predictable schedule.
What does FSRS stand for?
FSRS is the Free Spaced Repetition Scheduler, an open-source algorithm trained on hundreds of millions of real flashcard reviews. It schedules each review for the moment your predicted recall probability drops to your target retention.
Will switching to FSRS lose my Anki progress?
No. FSRS can use your existing review history when you enable it in Anki, and the change is reversible. You do not reset your cards or start over.
What target retention should I use with FSRS?
90% is a sensible default. Higher retention means more reviews but fewer forgotten cards; lower retention means fewer reviews but more lapses. Tune it to how high-stakes the material is.
Does Flica use FSRS or SM-2?
Flica uses FSRS by default, enabled and tuned out of the box, so you get the modern algorithm without any configuration.
The Bottom Line: FSRS Is the Better Default
SM-2 earned its place over decades and is still serviceable, but FSRS is the more efficient, more adaptive scheduler for the vast majority of learners. Same retention, meaningfully fewer reviews, that compounds over a long course.
If you don't want to manage algorithms at all, Flica ships FSRS on by default and adds AI card generation on top, free on iOS and Android. Build a deck from a YouTube lecture and let the modern scheduler do the timing.
Get FSRS Without the Configuration
Flica runs FSRS by default and turns YouTube, PDFs, and text into flashcards automatically, free on iOS and Android.
References
- Ye, J. (2023). Free Spaced Repetition Scheduler (FSRS), open-source repository and benchmarks
- Wozniak, P. A. (1990). Optimization of learning, SuperMemo SM-2 algorithm
- Anki Manual, FSRS section (accessed May 2026)
- Ebbinghaus, H. (1885). Memory: A Contribution to Experimental Psychology
- Settles, B. & Meeder, B. (2016). A Trainable Spaced Repetition Model for Language Learning. ACL