
The exact number of classic Sudoku solution grids is 6,670,903,752,021,072,936,960. That is the accepted count for completed 9×9 grids obeying the usual row, column and 3×3 box rules. It is not the same as the number of published puzzles.
The count comes from a computer-aided enumeration by Bertram Felgenhauer in 2005, independently checked by Frazer Jarvis. There is no Guinness-style official record for “most Sudoku grids counted”. This is a mathematical result, not a trophy, and it is the number mathematical references use when they discuss how many sudoku puzzles exist or the number of sudoku grids.
The confusion starts because “Sudoku combinations” can mean several different things. A finished grid, a puzzle with clues removed, a puzzle considered the same after swapping digits, and a puzzle considered the same after rotating the board are all different counting problems. The huge number above answers one clean question: how many completed classic Sudoku grids exist before you start treating lookalikes as the same?
So how many Sudoku grids are there, exactly?
There are 6,670,903,752,021,072,936,960 completed classic Sudoku grids. Write it in scientific notation and it is about 6.67 × 10²¹. Say it aloud and it is 6.67 sextillion, using the short scale.
This is the answer to a precise question: fill a standard 9×9 Sudoku grid with the digits 1 to 9 so that every row, every column and every 3×3 box contains each digit once. Count every finished grid that satisfies those rules. Do not remove clues. Do not ask whether it is hard. Do not merge grids that look alike after rotating, reflecting, or renaming the digits.
That last sentence does a lot of work. If you swap every 1 for a 7 and every 7 for a 1, the logical structure of the grid has not changed, but the written grid is different. The headline total counts it as different. The same is true if you permute rows within a band, columns within a stack, or make other allowed transformations.
Ask for a grid and you get the exact count. Ask for a clue puzzle and you must first say whether rotations, digit swaps and different clue sets are being counted separately. That is why the number of sudoku grids is stable, while puzzle totals shift with the definition.
Who proved the total, and on what date?
The result is attributed to Bertram Felgenhauer, with an independent check by Frazer Jarvis, in May 2005. There is no official record body for this count, and the surviving public references do not make it a dated award in the way a speed record would be. The best documented record is the mathematical result itself: named authors, named month, named year, exact total.
Felgenhauer’s calculation reduced the brute-force problem by fixing part of the grid, using symmetry, then counting the possible completions without listing every finished Sudoku one by one. Jarvis, a mathematician at the University of Sheffield, checked the enumeration independently and published an explanation of the method. That matters. A number this large is not persuasive because somebody printed it; it is persuasive because an independent enumeration reached the same total.
The factorisation of the total is exact:
6,670,903,752,021,072,936,960 = 9! × 72² × 2⁷ × 27,704,267,971
That does not make the count easier to picture, but it shows the answer is not a rounded estimate. The phrase “Sudoku combinations” can sound vague; this result is not vague. If a source treats the total as a rough guess, it is describing the completed-grid result incorrectly.
What counts as a distinct grid, and why does that matter?
The same finished Sudoku can be counted several ways. The right count depends on the question you are asking.
| Counting rule | What it treats as different | Best known figure | Why it matters |
|---|---|---|---|
| Completed written grids | Every valid finished grid with its actual digits in actual cells | 6,670,903,752,021,072,936,960 | This is the standard answer to the number of sudoku grids. |
| Digit renaming allowed | Grids that differ only by swapping symbols, such as all 1s and 7s | Smaller than the full grid count | Useful if you care about structure rather than printed digits. |
| Full Sudoku symmetries allowed | Renaming digits, rotating, reflecting, and legal row or column rearrangements | 5,472,730,538 essentially different grids | This is the common “essentially different” solution-grid count. |
| Clue puzzles | Partial grids that have exactly one completion | No single official total | This is what most players mean by a Sudoku puzzle. |
| Minimal clue puzzles | Puzzles where removing any given clue destroys uniqueness | Depends on the equivalence rules used | This is a stricter publishing and research category. |
This is where many wrong answers begin: they put a clue-puzzle question into a solution-grid box, then present the result as if nothing changed.
Why do people argue about smaller counts and 'different' puzzles?
People argue because Sudoku has several honest counting problems hiding under one casual question. If two puzzles have the same solution grid but reveal different clues, a player sees two different puzzles. If one puzzle is a rotated version of another, a mathematician may call them the same. Neither person is being silly. They are counting different things.
The best-known smaller solution-grid figure is 5,472,730,538 essentially different completed grids, attributed to Ed Russell and Frazer Jarvis in 2006. That count collapses grids that are equivalent under the usual Sudoku symmetries, including relabelling digits and rearranging rows or columns in permitted ways. It answers a structural question: how many genuinely different solution patterns are there?
For clue puzzles, the disagreements usually come from five choices:
- A puzzle is not a completed grid. It is a set of given clues with one valid completion.
- The same solution grid can support many different clue sets.
- Two clue sets can feel different even when one is a symmetry of the other.
- Difficulty is not determined by clue count alone.
- There is no single official record for the total number of publishable Sudoku puzzles.
The famous minimum-clue result is separate from the grid count. Gary McGuire, Bastian Tugemann and Gilles Civario proved that no standard Sudoku with 16 clues can have a unique solution; their paper appeared in 2012. That tells us 17 clues are necessary, but it does not give a grand total for every possible puzzle.
What does that number mean for real play, not just trivia?
- You will not run out of grids. The completed-grid count is so large that a puzzle site does not need to recycle finished solutions because the mathematical space is small. Repetition, when it happens, is a design or generation choice, not a shortage.
- A big grid count does not guarantee a good puzzle. A playable Sudoku needs a unique solution and a clue pattern that leads somewhere. A finished grid can be stripped of clues in bad ways, producing either multiple solutions or a dull solving path.
- Difficulty comes from technique. Two puzzles with the same number of clues can demand different work. One may fall to singles; another may need locked candidates, pairs, or more advanced reasoning. Clue count is visible, but it is a rough measure.
- Symmetry changes counts, not play. If a puzzle is rotated, reflected, or has its digits renamed, the solving logic is essentially the same. A player may still experience it as a fresh page because the eye has to read it anew.
- The exact total is a sanity check. If a page claims a wildly different number of classic completed grids without explaining symmetries, clue sets, or variants, it is probably mixing categories. The proven total is the anchor.
Frequently asked questions
No. A completed grid is an answer sheet, not a puzzle. A playable puzzle starts with some cells filled in and must force one solution. Many different clue patterns can point to the same finished grid, and many clue patterns are useless or ambiguous.
81 cells make up a standard Sudoku grid: 9 rows, 9 columns, and 9 smaller 3×3 boxes. Each row, column, and box must contain the digits 1 to 9 once. The huge grid count is the number of completed layouts that obey those three rules.
The common misconception is that fewer clues mean a harder Sudoku; the correction is that clue placement matters more than clue count. A puzzle with many givens can still require a difficult chain. A puzzle with fewer givens may fall quickly if the right singles and pairs appear early.
In a Klondike Solitaire (Turn 1) deal, uncovering a face-down card can change the line of play, even though the deck order was fixed from the start. Sudoku has the same split: one fixed finished grid can support many different starting clue patterns, some fair, some dull, some not valid puzzles at all.
It depends on uniqueness: if a puzzle must have exactly one solution, a standard 9×9 Sudoku needs at least 17 given cells. The known proof rules out every 16-clue puzzle. More clues do not guarantee an easier solve; placement and the required techniques matter more.
Sudoku X is stricter than standard Sudoku because the two main diagonals must also contain 1 to 9. That extra rule cuts down the possible completed grids and changes solving logic. Mini Sudoku (4×4) changes the size as well, so its count is a separate question, not a scaled-down copy.
Relabel digits and the grid is still valid; the exception is the counting question you are answering. If you count exact filled grids, swapping 1s and 2s creates another grid. If you count essentially the same pattern, it belongs with the original under symmetry.
Most people go wrong by treating each row, column, or box as if it can be filled independently. It cannot. A number placed in one cell constrains its row, its column, and its box at once. Those overlapping restrictions are exactly why the final count needed careful case reduction and checking.
An equivalence class is a group of grids treated as the same because one can be turned into another by allowed symmetries. Those symmetries include relabelling digits, swapping rows inside a band, swapping columns inside a stack, and rotating or reflecting the whole grid. Change the allowed symmetries and the count changes.
Choose a puzzle by the kind of solving you want, not by the size of the global grid count. On Veena Games, Sudoku is graded by the technique it needs, not by clue count. If you want a gentler start, use Easy Sudoku. If you want pressure, try Hard Sudoku or Expert Sudoku.
Nobody knows precisely, because “good” is not a mathematical condition. The best firm figure nearby is 6,670,903,752,021,072,936,960 completed standard grids. Each one can generate many clue patterns, but most possible patterns are not useful human puzzles: they may be ambiguous, trivial, or ugly to solve.
In 2005, the accepted total was obtained by a computation that used structure, not blind guessing. Symmetries cut the work down, then the remaining cases were counted and checked. That matters because a naïve search over all filled 9×9 arrays would be impossibly wasteful.
The cost of getting it wrong is a broken puzzle. If two different completed grids fit the same clues, logic cannot tell you which answer the setter intended. You may still fill a valid grid, but you are no longer solving a fair Sudoku. You are guessing between possible worlds.
Final thoughts
The huge number is fun, but the better lesson is smaller: Sudoku is precise. Change one word in the question and the answer changes. “Grid” means a completed solution. “Puzzle” means clues. “Different” might mean different ink on the page, or it might mean different mathematical structure.
That precision is part of the pleasure. A good Sudoku never asks you to guess, and a good count should not ask you to guess either. Felgenhauer and Jarvis gave the grid question a clean answer in 2005. Other questions are still wrapped in definitions: which symmetries are ignored, whether non-minimal puzzles count, whether two clue layouts with the same solution are separate entries.
If a friend asked me what to remember, I would give them the big number once, then tell them not to worship it. The count proves the supply is vast. It does not tell you which puzzle is elegant, fair, mean, or memorable. That comes from the path between the givens and the solution, one forced digit at a time.