FSRS vs SM-2 is a choice between two different scheduling models. SM-2 adjusts each card's interval with a compact, fixed formula. FSRS estimates difficulty, stability, and retrievability, then schedules toward a chosen retention target. FSRS gives modern Anki users more control and can fit parameters to review history; SM-2 remains easier to understand and implement.
Short answer: Use FSRS in a current version of Anki unless you need compatibility with a legacy scheduler or another app that only supports SM-2. If your deck has little history, FSRS can use its default parameters; you do not need to wait months before enabling it. SmartRecall currently uses SM-2, not FSRS.
FSRS vs SM-2 at a glance
| Question | SM-2 | FSRS |
|---|---|---|
| Core model | Ease factor and interval multipliers | Difficulty, stability, and retrievability |
| Primary input | Recall grade updates a fixed formula | Recall grade updates a trainable memory model |
| Retention target | Indirect | Explicit and configurable |
| Personalization | Per-card ease factor | Parameters can be optimized from review history |
| Fresh collection | Works immediately | Works with default parameters |
| Implementation | Small and predictable | More complex; scheduler plus optional optimizer |
| Current Anki support | Available as the legacy scheduler | Built in and documented by Anki |
| SmartRecall support | Yes | No |
This page focuses narrowly on FSRS vs SM-2. For a wider comparison that also covers Leitner, Anki's legacy scheduler, and physical-card workflows, read Spaced Repetition Algorithms Compared.
How SM-2 schedules reviews
SM-2 comes from Piotr Wozniak's early SuperMemo work. Each card keeps a small amount of state: an easiness factor, an interval, and a repetition count. A successful recall increases the interval; a failed recall shortens or resets it. The exact update depends on the grade.
That simplicity is the main advantage:
- The interval is easy to inspect and explain.
- The implementation is compact.
- The same inputs produce predictable results.
- Apps can support it without an optimizer or a large review log.
The limitation is that SM-2 does not explicitly estimate the probability that you can recall a card today. Its constants and interval updates are fixed. The easiness factor adapts per card, but the overall model does not learn a separate forgetting curve from your collection.
If you want the formula and a worked example, see How SM-2 Works.
How FSRS schedules reviews
FSRS stands for Free Spaced Repetition Scheduler. Its model tracks three concepts:
- Difficulty: how hard the card is to learn.
- Stability: how long the memory is expected to remain strong.
- Retrievability: the estimated probability of recalling the card now.
After a review, FSRS updates the card's memory state and chooses an interval that aims at the desired retention configured for the preset. Higher desired retention means shorter intervals and more reviews; lower desired retention means longer intervals and fewer reviews.
The optimizer is separate from the scheduler. The scheduler can use population-level default parameters immediately. When enough useful review history exists, Anki can optimize parameters for the cards covered by a preset. The current FSRS tutorial notes that Anki 24.06 and newer no longer impose a minimum review count for optimization; Anki decides which parameters it can fit from the available history.
The open-source SRS Benchmark compares FSRS versions, SM-2, and other models on large anonymized Anki review datasets. It measures prediction calibration and discrimination rather than promising a universal percentage reduction in study time. That distinction matters: better recall prediction does not guarantee the same time savings for every deck or learner.
The five differences that matter
1. Fixed interval rules vs a memory model
SM-2 asks: given the last interval, ease factor, and grade, what multiplier should be applied next?
FSRS asks: given this card's estimated memory state and the user's desired retention, what interval should come next?
Both respond to recall feedback. FSRS simply represents more of the memory state explicitly.
2. Ease factor vs difficulty and stability
SM-2 compresses much of a card's history into its ease factor. FSRS separates how difficult the item is from how stable the memory has become. Two cards can therefore have similar difficulty but very different stability, or vice versa.
3. Implicit vs explicit retention
SM-2 does not expose a direct desired-retention control. FSRS does. In Anki, desired retention is one of the most important FSRS settings. The official tutorial describes 80–95% as a reasonable range and 90% as a practical starting point for many users.
Do not raise retention automatically because a larger number sounds better. Higher retention produces more reviews. The useful setting is the one that matches the importance of the material and the time available.
4. Default parameters vs optimized parameters
FSRS personalization is optional. A new collection uses default parameters; an established collection can optimize them from its own history. This corrects a common misconception that FSRS is unusable until a learner has completed a fixed number of reviews.
Optimization is not continuous magic running after every sync. In current Anki workflows, the user runs Optimize or Optimize all presets, reviews the evaluation, and saves the preset. The official FSRS documentation remains the source of truth because this workflow changes as Anki evolves.
5. Compatibility and operational complexity
SM-2 is easier to port between products and easier to debug. FSRS requires more state, current formulas, and careful integration. Anki users also need to check add-ons that alter scheduling or intervals, because those can conflict with FSRS.
For an app choosing a scheduler, this operational difference can matter as much as benchmark performance. A correct SM-2 implementation is better than a partially implemented or inaccurately described FSRS integration.
Does Anki use FSRS or SM-2 now?
Current Anki documentation presents FSRS as its modern spaced-repetition algorithm while retaining a legacy SM-2-based scheduler. Anki integrated FSRS in version 23.10, and its manual now directs users toward FSRS.
For current Anki releases:
- Open the deck options for the relevant preset.
- Confirm that learning and relearning steps follow the current FSRS guidance.
- Set desired retention deliberately; 90% is a common starting point, not a universal optimum.
- Use Optimize when you want parameters fitted to available review history.
- Save the preset and evaluate the resulting workload over time.
Use Again when recall fails. The FSRS tutorial explicitly warns that pressing Hard after a failed recall tells the model that recall succeeded, which can produce intervals that are too long.
When to choose FSRS
FSRS is usually the better fit when:
- You use a current version of Anki.
- You want an explicit desired-retention setting.
- You have review history that can personalize the model.
- You are comfortable evaluating presets and scheduler settings.
- Your add-ons do not override scheduling behavior.
A new deck is not a reason to avoid FSRS. Default parameters are available for that case.
When SM-2 remains a reasonable choice
SM-2 remains reasonable when:
- Your app does not support FSRS.
- You need a transparent scheduler that is easy to reproduce.
- You are maintaining compatibility with an existing SM-2 collection.
- Your review workflow is small enough that scheduler tuning is not the bottleneck.
- You prefer fewer moving parts over model personalization.
Algorithm choice is also secondary to card quality and consistent retrieval practice. A sophisticated scheduler cannot rescue vague prompts, overloaded cards, or passive rereading. See Active Recall vs Passive Review for the learning behavior both schedulers depend on.
What SmartRecall uses
SmartRecall currently uses SM-2. It does not offer FSRS, FSRS parameter optimization, or side-by-side FSRS scheduling. Its verified workflow is different: users turn supported source material into editable flashcards, check the generated cards, and review them on an SM-2 schedule.
That boundary is important when comparing tools. Choose Anki when FSRS is a requirement. Choose SmartRecall when reducing manual card creation from PDFs or notes matters more, then review and edit the generated cards before studying. You can inspect the PDF-to-flashcards workflow before deciding.
Bottom line
The practical FSRS vs SM-2 decision is straightforward:
- For current Anki users, start with FSRS and its documented defaults.
- For products that only support SM-2, SM-2 remains a valid, understandable scheduler.
- Do not switch based on a universal percentage claim; compare retention, workload, and consistency on your own collection.
- Do not treat the scheduler as a substitute for well-formed cards and honest recall grading.
For the broader landscape, continue with SM-2, FSRS, Leitner, and Anki compared.
FAQ
Is FSRS better than SM-2?
FSRS has a more expressive memory model, an explicit desired-retention setting, and optional parameter optimization from review history. Public benchmarks show stronger recall-prediction performance than SM-2, but that does not translate into one guaranteed time-saving percentage for every learner. In a current Anki version, FSRS is the sensible default; SM-2 remains useful when compatibility or implementation simplicity is the priority.
Does FSRS need 1,000 reviews before it works?
No. FSRS can schedule with default parameters from the beginning. Review history is used to optimize parameters for a collection. The current FSRS tutorial says Anki 24.06 and newer no longer enforce a minimum review count for optimization; Anki selects which parameters can be optimized from the data available.
What desired retention should I use with FSRS?
The official FSRS tutorial describes 80–95% as a reasonable range and 90% as a practical starting point for many users. A higher target creates more reviews and shorter intervals. Choose deliberately based on the importance of the material and your sustainable workload instead of assuming that the highest possible retention is always best.
Does Anki use FSRS by default?
Modern Anki documentation uses FSRS as its current scheduling model and keeps the older SM-2-based scheduler as a legacy option. Because defaults and migration behavior can change between releases, check the current Anki manual and the deck preset you are actually using rather than relying on an older tutorial.
Does SmartRecall use FSRS?
No. SmartRecall currently schedules reviews with SM-2. It does not claim to provide FSRS optimization or FSRS scheduling. Its distinct feature is turning supported PDFs or pasted study material into editable flashcards that users should review before studying.
