16×16 Sudoku (Hexadoku): Rules, Tips & Notation

16×16 sudoku is the same puzzle with the volume turned up. Instead of nine rows, nine columns and nine boxes, you get sixteen of each, and instead of the digits 1 to 9 you get sixteen distinct symbols. The rule is unchanged in spirit: every row, every column and every 4×4 box must contain each symbol exactly once.

You will see it sold as hexadoku, hexadecimal sudoku, Sudoku 16, or simply 16×16. All of those names describe the same grid. What differs between publishers is the symbol set, and that detail matters more than it sounds — it is the single most common reason a copied puzzle fails to import into a solver.

The rule change, in numbers

A standard sudoku has 81 cells, 27 units and nine symbols. A 16×16 grid has 256 cells, 48 units and sixteen symbols.

The symbol set is usually the digits 1 to 9 followed by the letters A, B, C, D, E and F, with G as the sixteenth. That is seven letters, which combined with nine digits gives sixteen. Some publishers instead use a hexadecimal set of 0 to 9 plus A to F, and a few use shapes or colours. The rules are identical in every case; only the labels change.

One consequence deserves to be stated plainly, because it is the thing that most often confuses newcomers: the symbols are labels, not numbers. Nothing in a standard 16×16 puzzle depends on arithmetic. A is not "worth ten" in any solving sense, and you will never add two cells together. That is why substituting letters for digits, or colours for letters, is a legitimate solving technique rather than a cheat.

Why it is much harder than 9×9

Players expect a bigger grid to take maybe twice as long. It does not, and the reasons are structural rather than psychological.

The candidate lists grow. A 9×9 cell can hold at most nine candidates; a 16×16 cell can hold sixteen. Pencil marks that fit comfortably in a small cell on paper become unreadable on screen, and the sheer number of marks per cell makes visual scanning slower.

Naked singles become rare. A naked single needs fifteen of the sixteen symbols already placed in a cell's row, column and box combined. In a 9×9 grid you need eight placements to force a cell; here you need fifteen, and the grid has four times as many cells competing for those placements. Early in a 16×16 solve you will place almost nothing by naked singles.

Every sweep costs four times as much. Cross-hatching a single digit means checking sixteen boxes instead of nine, and each box is 4×4 rather than 3×3. The work per digit roughly quadruples, and there are sixteen digits rather than nine.

Working memory fails. A 9×9 grid can be held in your head in chunks. A 16×16 grid cannot. You need external notation, and you need it from the first minute rather than after you get stuck.

Errors propagate further and cost more. Placing a wrong digit in a 9×9 grid usually reveals itself within a few moves. In a 16×16 grid, a wrong digit can sit quietly for twenty deductions before the contradiction surfaces, and by then you cannot tell which of the twenty was the mistake. This is why checkpointing, described below, matters more than raw technique.

Symbol substitution: make the grid easier to see

If the letters slow you down, change them. The puzzle does not care what the symbols are, so any consistent relabelling is valid.

Map letters to two-digit numbers. A becomes 10, B becomes 11, and so on to G as 16. This works well if you think numerically, but it makes the grid visually denser and the cells harder to scan. It is best on paper, where you can write the conversion once at the top.

Map symbols to colours. Sixteen distinct colours is a lot, but you do not need sixteen. Colour the seven letter-symbols and leave the digits alone, so that the grid separates into "number cells" and "colour cells". This makes letter clusters visible at a glance, which is exactly what cross-hatching needs.

Map to a smaller alphabet of shapes. Squares, circles, triangles and so on, seven of them, for the same reason. The advantage over colours is that shapes survive greyscale printing and most forms of colour blindness.

Whatever you choose, write the legend down and never change it mid-solve. The classic failure is to relabel halfway through, which silently converts every earlier deduction into a guess. If you catch yourself thinking "wait, was D thirteen or fourteen", stop and rebuild from the legend.

How to scan a 16×16 grid without losing your place

Scanning is the skill that separates a pleasant two-hour solve from a four-hour slog. Four passes do most of the work, and they should be done in this order.

Pass 1 — the symbol sweep

Take one symbol and find every place it already appears. Then, for each of the sixteen boxes, ask whether that symbol can still go there. In a 16×16 grid each symbol must appear sixteen times, once per row, once per column, once per box. Working one symbol at a time across the whole grid is slower per pass than working one box at a time, but it produces far more placements, because a symbol's sixteen instances constrain each other.

Do this for the symbols that are already most numerous. A symbol with six or seven givens will often fall into place in several boxes during a single sweep; a symbol with one given will not, and you should not spend time on it yet.

Pass 2 — the band sweep

A band is four consecutive rows. In a 16×16 grid there are four bands, and each band contains four boxes side by side. Within a band, each symbol appears exactly four times: once in each row, and once in each of the four boxes. That double requirement is much tighter than it looks, and it is the fastest way to place digits in a big grid.

Work band by band, and inside each band work symbol by symbol. For a given symbol and a given box in the band, ask which of the box's four rows can take it. If only one row can, you have a placement — or at minimum a line to eliminate from.

Pass 3 — the stack sweep

A stack is four consecutive columns, and it mirrors the band sweep exactly: four boxes stacked vertically, each symbol once per column and once per box. Bands and stacks catch different things, so do both rather than assuming one implies the other.

Pass 4 — the box audit

Finally, go box by box and count. A 4×4 box holding nine or ten symbols is close to resolving; a box holding three is not worth attention yet. Sort the boxes by how full they are and work the fullest first. This ordering habit alone saves a great deal of wasted scanning.

Throughout all four passes, keep a physical or on-screen pointer. On paper, a fingertip or a straightened paper clip works. On screen, use the browser's find function or the solver's highlight feature. The most common scanning error in a 16×16 grid is not a logic mistake; it is looking at the wrong row.

The box-in-a-box relationship

The sixteen 4×4 boxes are themselves arranged as a 4×4 grid. That second-level grid is the mental model that makes large puzzles tractable, and it is worth building explicitly.

Number the boxes row-major: B1 to B4 across the top band, B5 to B8 across the second band, and so on to B16. Box B6 then sits in band 2 and stack 2, covering rows 5 to 8 and columns 5 to 8.

Three relationships follow, and each has a use.

• A box shares its band with three other boxes. Those four boxes together cover four rows and all sixteen columns, and between them they must place each symbol exactly four times. If a symbol is already placed twice in a band, only two placements remain for that band, and the boxes that can host them are often forced.
• A box shares its stack with three other boxes. The same argument runs vertically. Bands and stacks are not redundant: a symbol can be constrained horizontally and free vertically, and the intersection of the two sets is where placements appear.
• A cell belongs to exactly one box, one band and one stack. So a candidate can be eliminated by a box-level reason, a band-level reason or a stack-level reason, and experienced solvers check all three before writing a candidate down.

This is the "box in a box" idea in its useful form: the small 4×4 box is the unit you fill, and the 4×4 arrangement of those boxes is the unit you navigate. When you are lost in a 16×16 grid, the recovery move is always the same — zoom out, pick a band, and ask where one symbol can go within that band.

How long a 16×16 actually takes

Time expectations are personal, so treat any hard number with suspicion. What can be said reliably is the shape of the effort.

A 9×9 puzzle is a single sitting of minutes. A 16×16 puzzle is a project measured in hours, and your first one will almost certainly take more than one sitting. That is normal and is not a sign that you are doing it wrong.

The difficulty curve inside a single 16×16 solve is also unusual. The opening is fast, because the given symbols support plenty of cross-hatching. The middle is slow, because you are down to candidates and long chains of elimination. The endgame accelerates again as boxes fill and naked singles reappear. Plan your session around that shape: start when you have a clear hour, and stop at a natural checkpoint rather than in the middle of a chain of deductions.

If you are choosing a first 16×16, prefer one with a generous sprinkling of givens in every band. A grid with empty bands is not a harder puzzle in an interesting way; it is a longer one.

Getting a 16×16 into a solver correctly

Most solver failures on large grids are input failures, not logic failures. Four checks catch nearly all of them.

Check that the solver supports 16×16 at all. Many sudoku solvers are hard-wired to 9×9 and will either reject the input or silently truncate it. Our own sudoku solver is built for the standard grid, so for a 16×16 you want a tool that explicitly offers a size selector.

Match the symbol convention. This is where copied puzzles break. In the common 1-to-9-plus-A-to-G convention, A is the tenth symbol and G is the sixteenth. In the hexadecimal 0-to-9-plus-A-to-F convention, 0 is the first symbol and F the sixteenth, which makes A the eleventh. Paste one into the other without remapping and every cell from the tenth onward shifts by one position. The puzzle will import without complaint and refuse to solve, and you will spend an hour looking for a logic error that does not exist.

Get the order right: row by row, left to right. Almost every solver expects 256 characters in row-major order, with blanks entered as 0 or a full stop. Sixteen characters per row, sixteen rows. If the tool accepts newlines, use them; they make a mis-typed row obvious.

Verify after import. Count the givens before and after. Pick two cells from opposite corners and confirm the symbols match the original. If your puzzle has 60 givens, the imported grid should show 60 filled cells. Solvers that display a given count make this trivial, and it is worth choosing one that does.

Column-major entry is the other classic failure. If a solver returns "no solution" on a puzzle you know is valid, re-enter the first two rows by hand and compare. The error is almost always in the transposition, not in the puzzle.

Notation worth learning

Large-grid solving lives or dies on notation, because nobody can hold 256 cells in their head.

• Cell references. R5C12, or r5c12, means row 5, column 12. Coordinates run from the top-left. Some tools number from zero, so r0c0 is the top-left; check before you write a reference down.
• Box references. B1 to B16, numbered row-major. B6 is the second box in the second band.
• Band and stack. Band 1 is rows 1 to 4, band 2 is rows 5 to 8, and so on. Stack 1 is columns 1 to 4. Some writers use "chute" for either a band or a stack, inherited from 9×9 terminology where a chute is three rows or three columns.
• Candidate sets. Braces around the remaining possibilities, as in {3, 7, B, F}. Sort them so that scanning your own notes is fast.
• The row string. A 16-character line such as 53..7....A..9.F2, used for solver input. Sixteen of these lines make a grid.
• Placement versus elimination. Keep the two visually distinct in your notes. A digit you have written as a candidate and a digit you have committed to the grid must never look the same, or you will build on a guess.

The intermediate steps: 6×6 and 12×12

Jumping from 9×9 to 16×16 is a bigger step than most players expect. The 6×6 and 12×12 grids exist precisely to bridge it, and both are worth a session.

6×6 uses the symbols 1 to 6 and boxes of 2 rows by 3 columns, giving six boxes in a 2-by-3 arrangement. The box shape is the interesting part: because it is not square, a box constrains two rows and three columns rather than an equal span in each direction. That asymmetry is excellent practice for the 12×12, where boxes are also rectangular.

12×12 uses twelve symbols, usually 1 to 9 then A, B and C, with boxes of 3 rows by 4 columns. Twelve boxes, arranged 3 by 4. This is the true rehearsal for hexadoku: the grid is large enough that you must scan in bands, the symbol set is mixed digits and letters, and the boxes are rectangular so the box-band relationship has to be reasoned about rather than assumed. It is also the size at which pencil marks stop fitting comfortably on paper.

A sensible progression is 6×6 for a single session to internalise rectangular boxes, then 12×12 until band scanning feels automatic, then 16×16. Players who skip the 12×12 usually report that the 16×16 feels chaotic rather than merely long, and the missing skill is band reasoning.

If you want a gentler start in the other direction, our 6×6 mini sudoku is the easiest rectangular-box grid on the site, and the core strategy guide covers the techniques that carry over unchanged.

The difficulty ladder

The jump from 12 to 16 is smaller than the jump from 9 to 12, because by then the habits are in place. The jump from 16 to 25 is mostly a test of stamina and of whether your notation system still works when cells get small.

Techniques that scale, and one that does not

Every classic technique survives the move to sixteen symbols, because none of them depend on the symbols being digits. Naked singles, hidden singles, naked and hidden pairs and triples, locked candidates, X-wings, swordfish, XY-wings, colouring and chains all work exactly as before, with sixteen-symbol candidate sets instead of nine.

What changes is the cost of applying them. A hidden single in a 16×16 grid requires checking sixteen cells in a unit rather than nine, and each check involves a larger candidate set. Fish techniques that felt routine on 9×9 become expensive, not because the logic is harder but because the visual search is. In practice most solvers lean heavily on scanning and hidden singles up to the middle game and only reach for fish when the grid forces it.

The technique that does not scale is mental arithmetic. There is no sum, no difference, no parity relationship between symbols in a standard hexadoku. If you find yourself reasoning about A plus B, you have imported a habit from killer sudoku or from XV sudoku, and it does not apply here. Sudoku 16 is pure combinatorics of placement.

Common mistakes

1. Mismatched symbol conventions. Pasting a 1-to-9-plus-A-to-G grid into a tool expecting 0-to-9-plus-A-to-F shifts everything from the tenth symbol onward. The puzzle imports and refuses to solve.
2. Entering columns as rows. The other silent import failure. Verify a couple of rows by hand after importing.
3. Using 3×3 boxes out of habit. The box here is 4×4. A solver configured for 9×9 will produce nonsense, and a human working from a 3×3 mental model will place wrong digits confidently.
4. Counting a band as three rows. Bands are four rows in this grid. The 9×9 habit of three-row chutes is the most persistent error in the whole variant.
5. Working without written candidates. On 9×9 you can get away with it. Here, working memory gives out somewhere in the middle game and you start re-deriving the same eliminations.
6. Never checkpointing. Because errors propagate silently, save or photograph the grid at each clean milestone. When a contradiction appears, you can roll back to the last checkpoint instead of rebuilding from scratch.
7. Changing the symbol mapping mid-solve. Relabelling halfway invalidates every earlier deduction. Fix the legend at the start.
8. Assuming a bigger grid means a harder puzzle. A 16×16 with generous givens is long but gentle. Difficulty comes from the solve path, not from the cell count.

Frequently asked questions

What is 16×16 sudoku?

It is a sudoku on a 16 by 16 grid, also called hexadoku or hexadecimal sudoku. Every row, every column and every 4 by 4 box must contain sixteen distinct symbols exactly once. Most versions use the digits 1 to 9 followed by the letters A to G, giving 256 cells and 48 units in total.

Why is 16×16 sudoku so much harder than 9×9?

Candidate lists grow from nine to sixteen, so naked singles are far rarer, and every cross-hatching sweep costs about four times as much work. The grid is also too large to hold in your head, so you depend on written notation, and a wrong digit can survive twenty deductions before it surfaces.

Can I use letters or colours instead of numbers?

Yes. The symbols are labels rather than numbers, and nothing in a standard 16 by 16 puzzle depends on arithmetic, so any consistent relabelling is valid. Mapping the seven letter-symbols to colours or shapes makes clusters easier to spot, provided you write the legend down and never change it mid-solve.

How do I enter a 16×16 puzzle into a solver?

Use 256 characters in row-major order, sixteen per row, with 0 or a full stop for blanks, and make sure the solver's symbol convention matches the puzzle. The common 1 to 9 plus A to G set treats A as the tenth symbol, while the hexadecimal 0 to 9 plus A to F set treats A as the eleventh, so a careless paste shifts every symbol from the tenth onward.

What are bands and stacks in a 16×16 grid?

A band is four consecutive rows and a stack is four consecutive columns. Each band contains four boxes side by side and each stack contains four boxes stacked vertically. Every symbol must appear exactly four times in each band and four times in each stack, which makes band and stack sweeps the fastest way to find placements.

Should I start with 12×12 before trying 16×16?

Yes, if the 16 by 16 feels chaotic rather than merely long. A 12 by 12 grid uses 3 by 4 boxes and twelve symbols, which rehearses rectangular boxes, mixed digit-and-letter symbols and band scanning at a size that is still manageable. Players who skip it often struggle with band reasoning rather than with logic.

Related: Play sudoku free · Sudoku solver · Mini sudoku (6×6) · Core solving strategies · Sudoku X and XV variants