Here is a number that might keep you up at night: there are more valid Sudoku grids than grains of sand on every beach on Earth. The humble 9×9 grid, governed by three simple rules, produces an almost incomprehensible number of unique configurations. The mathematics behind counting these grids is a fascinating story of combinatorics, symmetry, and computational brute force — and the final answer is staggering enough to guarantee you will never run out of puzzles.
The Big Number: 6.67 Sextillion
In 2005, mathematicians Bertram Felgenhauer and Frazer Jarvis published a landmark calculation that answered the question definitively for the standard 9×9 Sudoku grid. The total number of valid completed grids — where every row, column, and 3×3 box contains the digits 1 through 9 exactly once — is:
$$6{,}670{,}903{,}752{,}021{,}072{,}936{,}960$$
That is approximately $6.67 \times 10^{21}$, or 6.67 sextillion. To give you a sense of this magnitude:
| Comparison | Approximate Count |
|---|---|
| Seconds since the Big Bang | $4.3 \times 10^{17}$ |
| Grains of sand on Earth | $7.5 \times 10^{18}$ |
| Valid Sudoku grids | $6.67 \times 10^{21}$ |
| Stars in the observable universe | $2 \times 10^{23}$ |
| Atoms in a human body | $7 \times 10^{27}$ |
| Atoms in the observable universe | $\sim 10^{80}$ |
The number of Sudoku grids is about 890 times larger than the number of grains of sand on every beach and desert on the planet. It exceeds the number of seconds that have passed since the beginning of the universe by a factor of over 15,000.
Felgenhauer and Jarvis did not arrive at this number through a single elegant formula. The calculation required a clever combination of mathematical reasoning and computational enumeration. They broke the problem down by first counting the ways to fill the top band of three rows, then counting the compatible completions for the remaining rows. Even with significant mathematical simplification, the final step required exhaustive computer search.
How the Calculation Works
Understanding how mathematicians counted all valid grids reveals interesting aspects of Sudoku’s structure.
The Top Row Is Easy
The first row of a Sudoku grid can be any permutation of the digits 1 through 9. That gives us $9! = 362{,}880$ possibilities. By convention, though, mathematicians often fix the first row as 1, 2, 3, 4, 5, 6, 7, 8, 9 and multiply by $9!$ at the end, since every valid grid has exactly one version with each possible first row.
Constraints Multiply Rapidly
Once the first row is placed, the second row is heavily constrained. It cannot repeat any digit in the same column, and the first three cells of row two are further constrained by the first 3×3 box. As you fill more rows, the constraints compound. By the time you reach the bottom rows, there may be very few — or zero — valid completions.
The Band Approach
Felgenhauer and Jarvis divided the grid into three horizontal bands of three rows each. They first enumerated all valid configurations for the top band (rows 1–3), then for each top band configuration, they counted the number of valid middle and bottom bands. This divide-and-conquer approach made the computation feasible.
The number of valid completions of the top band alone (given a fixed first row) is 948,109,639,680. From there, the researchers computed how many ways the remaining six rows could be completed, accounting for all constraints.
Computational Verification
The result was verified independently by Ed Russell and Frazer Jarvis using a different computational approach, confirming the figure. This kind of independent verification is essential in computational mathematics — when the answer has 22 digits, you want to make sure you counted correctly.
Symmetry and Essentially Different Grids
The raw count of 6.67 sextillion includes many grids that are, in a meaningful sense, the “same” puzzle. Consider a completed Sudoku grid. You could:
Relabel the digits: Replace every 1 with a 5, every 5 with a 3, and so on. The grid follows the same logical structure with different labels. Since there are $9! = 362{,}880$ ways to relabel nine digits, each fundamentally distinct grid produces 362,880 relabeled variants.
Permute rows within a band: Swapping the second and third rows within the top band produces a different grid that follows the same constraints. Each band has $3! = 6$ internal row permutations, and there are three bands.
Permute columns within a stack: Similarly, columns within each three-column stack can be rearranged. That gives another $6^3$ permutations.
Permute the bands themselves: The three horizontal bands can be reordered ($3! = 6$ permutations), as can the three vertical stacks ($3! = 6$ permutations).
Transpose the grid: Reflecting the grid across its main diagonal (swapping rows and columns) produces another valid grid.
When you combine all these symmetry operations, you get a group of transformations with $2 \times 9! \times 6^8 = 1{,}218{,}998{,}108{,}160$ elements. Applying Burnside’s lemma (a tool from group theory that counts distinct objects under symmetry), the number of essentially different completed Sudoku grids is:
$$5{,}472{,}730{,}538$$
That is approximately 5.47 billion essentially different grids. While this is dramatically smaller than 6.67 sextillion, it is still an enormous number. You could solve one essentially different grid every day and not repeat for over 15 million years.
From Grids to Puzzles: An Even Bigger Number
Here is where things get really mind-bending. The numbers above count completed grids — fully filled Sudoku solutions. But a Sudoku puzzle is not a completed grid. A puzzle is a partially filled grid with enough clues to lead to exactly one solution. And the number of possible puzzles is vastly larger than the number of completed grids.
From a single completed grid, you can create many different puzzles by choosing which cells to reveal as clues. A 9×9 grid has 81 cells, so the number of possible subsets of cells is $2^{81} \approx 2.4 \times 10^{24}$. Not all of these subsets produce valid puzzles (most will not have a unique solution), but a substantial fraction of the viable subsets do.
Research suggests that the average completed grid can yield thousands to millions of distinct valid puzzles, depending on how many clues are given and the specific structure of the grid. This means the total number of possible Sudoku puzzles is conservatively in the range of $10^{25}$ or higher — hundreds of times more than the number of completed grids.
For more about how puzzles are generated from completed grids, see our article on how Sudoku puzzles are made.
The 17-Clue Minimum
One of the most celebrated results in Sudoku mathematics is the proof that 17 is the minimum number of clues needed for a valid puzzle with a unique solution.
Before this was proven, thousands of 17-clue puzzles had been discovered, but no valid 16-clue puzzle had ever been found despite extensive searching. The question of whether 16-clue puzzles existed became one of the most prominent open problems in recreational mathematics.
In 2012, Gary McGuire, Bastian Tugemann, and Gilles Civario settled the question definitively. Using a highly optimized algorithm running on a high-performance computing cluster, they exhaustively showed that no 16-clue Sudoku puzzle with a unique solution exists. The computation took approximately 7.1 million CPU-hours — equivalent to about 800 years on a single processor.
This result has practical implications for puzzle design. Every valid Sudoku puzzle must contain at least 17 revealed digits. The hardest puzzles tend to have clue counts near this minimum, though having more clues does not automatically make a puzzle easier — the positioning of clues matters enormously. Our evil puzzles often have relatively few clues, requiring advanced techniques to solve.
How Clue Count Affects Difficulty
While the minimum is 17, typical puzzles have more clues. Here is a rough guide:
| Clue Count | Typical Difficulty | Solving Approach |
|---|---|---|
| 17–20 | Extremely hard to evil | Advanced chains, coloring, trial-and-error often needed |
| 21–25 | Hard to expert | Named techniques like X-Wing, XY-Wing |
| 26–30 | Medium to hard | Naked pairs, hidden pairs, triples |
| 31–35 | Easy to medium | Singles dominate, occasional pairs |
| 36–45 | Easy | Naked singles and hidden singles only |
Note that this is a rough correlation, not a strict rule. A 28-clue puzzle with clues in a tricky configuration can be harder than a 24-clue puzzle with a more revealing clue pattern.
Why You Will Never Run Out
Let us put the numbers in perspective with a thought experiment.
Suppose you solve one Sudoku puzzle per minute, 24 hours a day, 7 days a week, without ever sleeping, eating, or resting. That gives you 525,600 puzzles per year, or roughly half a million.
At that grueling pace, to solve all 6.67 sextillion completed grids (not even counting the much larger number of possible puzzles), you would need:
$$\frac{6.67 \times 10^{21}}{525{,}600} \approx 1.27 \times 10^{16} \text{ years}$$
That is about 12.7 quadrillion years — approximately 920,000 times the current age of the universe (13.8 billion years). And that is just the completed grids. If you include all valid puzzles derivable from those grids, the time required becomes even more absurd.
The practical consequence is liberating: no matter how dedicated a solver you are, you will encounter new and unique puzzles for the rest of your life. Even the most prolific daily solvers have completed fewer than one hundred-trillionth of one percent of all possible puzzles.
Implications for Puzzle Generators
These astronomical numbers explain why computerized puzzle generators are so effective. When there are sextillions of valid grids to draw from, the chances of generating the same puzzle twice are effectively zero.
Modern puzzle generators like the one powering SudokuPulse typically work by:
- Generating a random valid completed grid using constraint satisfaction algorithms
- Systematically removing clues while verifying that the remaining puzzle has a unique solution
- Testing the difficulty by running a human-strategy solver to determine which techniques are needed
- Classifying the puzzle into the appropriate difficulty level
The vastness of the solution space means each new puzzle is genuinely fresh. Combined with the difficulty classification step, generators can produce an essentially infinite supply of quality puzzles at every level from easy to evil.
Sudoku Numbers in Mathematical Context
The mathematics of Sudoku has connections to several deep areas of mathematics:
Constraint satisfaction: Sudoku is a constraint satisfaction problem, a class of problems studied extensively in computer science and operations research. The number of valid grids is related to the number of solutions in the constraint model.
Latin squares: A Sudoku grid is a special type of Latin square — an $n \times n$ array filled with $n$ different symbols, each occurring exactly once in each row and column. The box constraint makes Sudoku a further-constrained variant. The number of $9 \times 9$ Latin squares (without the box constraint) is approximately $5.52 \times 10^{27}$, showing how much the box constraint reduces the solution space.
Graph coloring: As mentioned in our article on Sudoku and mathematics, solving Sudoku is equivalent to coloring a specific graph with 9 colors. The number of valid grids corresponds to the number of proper colorings of the Sudoku graph.
NP-completeness: While solving a specific Sudoku puzzle is straightforward (in practice) for a 9×9 grid, the generalized Sudoku problem ($n^2 \times n^2$ grid) is NP-complete. This means no algorithm is known that can solve all generalized Sudoku puzzles efficiently as the grid size grows, connecting Sudoku to one of the deepest open questions in computer science.
Beyond 9×9: Counting Larger Grids
The counting problem becomes even more extreme for larger grid sizes:
| Grid Size | Estimated Valid Grids |
|---|---|
| 4×4 | 288 |
| 9×9 | $6.67 \times 10^{21}$ |
| 16×16 | Unknown (estimated $> 10^{100}$) |
| 25×25 | Unknown (staggeringly large) |
For 16×16 Sudoku, the exact count has not been calculated. The computational resources required would far exceed current capabilities. Mathematicians estimate the number is at least $10^{100}$ (a googol), and likely much larger. For 25×25 grids and beyond, the numbers are so large that even estimating them is an active research challenge.
This exponential growth in grid size illustrates a fundamental property of combinatorial problems: small increases in the problem dimensions lead to enormous increases in the number of possibilities.
Frequently Asked Questions
How many valid completed Sudoku grids exist?
There are exactly 6,670,903,752,021,072,936,960 valid completed 9×9 Sudoku grids — approximately $6.67 \times 10^{21}$, or 6.67 sextillion. This was calculated by Bertram Felgenhauer and Frazer Jarvis in 2005.
How many essentially different Sudoku grids are there?
When you remove grids that are equivalent through symmetry operations like rotation, reflection, and relabeling of digits, there are approximately 5.47 billion essentially different Sudoku grids.
What is the minimum number of clues for a valid Sudoku puzzle?
The minimum is 17 clues. No valid Sudoku puzzle with a unique solution has ever been found with only 16 clues, and in 2012, Gary McGuire and colleagues proved computationally that 16-clue puzzles with unique solutions do not exist.
Will I ever run out of Sudoku puzzles to solve?
Effectively, no. Even if you solved one puzzle every second without stopping, it would take longer than the age of the universe to exhaust all possible Sudoku puzzles. You will never run out.
How does the number of Sudoku grids compare to atoms in the universe?
The observable universe contains an estimated $10^{80}$ atoms, which is vastly more than the $6.67 \times 10^{21}$ Sudoku grids. However, when counting all possible puzzles (not just completed grids), the numbers become astronomically large and approach cosmological scales.