A Wordle solver for the moment when three guesses are in and the list in your head has run out. Enter each guess you've made and set the tile colours to match the game — grey, yellow, green — and the tool narrows a dictionary of five-letter words to the ones still possible, counts them, and lists them with the words made of the most common remaining letters first. It is a filter and a sorter, not an oracle: it does not know today's word, and its dictionary is not the Times' answer list. The answer itself is on today's Wordle answer page and the staged clues on the hints page; the solver is for people who want to keep guessing, just better.
Try the wordle solver
Type a guess, tap each tile to set its colour (grey → yellow → green), then add it. The list narrows and re-ranks after every guess.
How the filter works
Every guess you enter is turned into a rule. A green tile fixes that letter in that position. A yellow tile says the letter is in the answer but not in that position. A grey tile says the letter is not in that position and, if the same letter got no green or yellow elsewhere in the guess, not in the word at all. If the same letter appears twice in a guess and only one tile is coloured, the solver reads the grey as "exactly one of this letter" — the same rule the game applies — which is why its list will sometimes keep a word you would have crossed out.
The rules are applied to a public-domain English word list (the dwyl english-words collection) cut down to 15,910 five-letter words. That list is not the Times'. It shares 10,625 words with the 14,855 guesses NYT's game code accepts, includes thousands of obscure words the game rejects, and is missing nine words from the original answer list — DETOX, GEEKY, INBOX, LATTE, PENNE, PESTO, RAMEN, REHAB and WIMPY. Every one of the 41 answers HintSheet published from 1 September to 11 October 2026 was on it.
How the ranking works
Once the list is filtered, the solver sorts it. For the surviving words it counts how often each letter appears in each of the five positions and how many words contain each letter at all, then scores every candidate by adding its letters' positional counts plus twice their overall counts, counting a repeated letter once. Words made of letters that are common among the survivors rise to the top; the first three are highlighted, and up to sixty are shown with a count of the rest.
That has two consequences worth knowing. The solver only ever suggests words that are still possible, so it never proposes a "probe" word that tests new letters but cannot be the answer. And because the dictionary is broad, the top of the list is often an obscure word: with no guesses entered, the first five are TARES, ARIES, LARES, RALES and NARES, and the game itself rejects ARIES. Prefer the common word when two are close.
Entering guesses accurately
The colours are the input; the word list does the rest.
- Type the guess exactly as you played it, then tap each tile until it matches the game's colour. New tiles default to grey.
- Double letters: colour each tile independently. The game colours them independently and so does the solver.
- A guess the game rejected as "not in word list" was never scored, so leave it out.
- Enter guesses in the order you played them; it makes no difference to the filter but keeps the display readable.
Starting words the solver recommends
With no information the solver's ranking is the same every day, and it is not a good opener list: it favours plural-looking words built from E, A, R, S and T, which the Times rarely uses as answers. For a first guess, use a word chosen against the actual answer list instead. We scored openers over the 2,309-word original answer list in NYT's game code.
| Opener | Why it scores well | Average answers left after guess 1 |
|---|---|---|
| ROATE | Fewest answers left on average of all 14,855 accepted words | 60.1 |
| RAISE | Nearly as low, and a word everyone knows | 60.7 |
| TRACE | Most distinct colour patterns of any accepted word (150) | 73.9 |
| SLATE | S first, E last; 146 distinct patterns | 71.3 |
| CRANE | Common letters in common positions | 78.7 |
About the word list
The Times bought Wordle in January 2022 and, according to Wikipedia, removed words it judged offensive or insensitive from the answer list; since November 2022 an editor, Tracy Bennett, has chosen each day's word. In the 41 answers from 1 September to 11 October 2026, 15 were not on the original 2,309-word list. The solver does not try to track the Times' choices. It filters its own 15,910-word list, which contained all 41 of those answers but is missing nine words from the original list (LATTE and RAMEN among them). If your answer might be one of those, the solver will not show it.
No solver can know a word the Times has not used yet. What this one promises is narrower: every word it shows is consistent with the colours you entered, and the count tells you how many such words its dictionary holds.
When to use it and when not to
The honest use is guess four onward. By then the information is mostly in and the remaining work is enumeration, which is exactly what people are bad at and software is good at. Using it on guess one or two is legitimate but joyless, and its suggestions at that stage are poor openers anyway. A middle path is to use it only to check that a planned guess is not a wasted one — enter your guesses, look at the count, and see whether your intended word is still on the list.
For learning rather than finishing, the Wordle guide covers starting words, second-guess strategy and the repeated-letter trap in more depth, and the Wordle archive lists every answer HintSheet has published.
Limits, stated plainly
Four things the solver does not do:
- It does not know today's answer. It filters and sorts; the answer page knows the answer.
- It does not use the Times' lists. Its dictionary is broader in most places and missing a few answers (LATTE, PENNE, RAMEN and six more).
- It does not suggest probe words or switch modes. Every suggestion is a word still consistent with your colours.
- It does not handle other word lengths. This tool is for the five-letter game.
Worked example: from guess three to the answer
A real board: the answer on 26 September 2026 was ELEGY. Open with SLATE and the game shows a green L in second place, a yellow E and three greys. Entered into the solver, that leaves 72 words. Follow with CRONY — five new letters — and the only colour is a green Y at the end. The solver is now down to six: GLUEY, GLEBY, BLUEY, GLEDY, FLUEY and ELEGY, in that order.
That list shows both what the tool is good at and its main weakness. Six candidates is a short enough list to finish in one or two guesses. But the real answer is sixth, behind five words that are rare or not accepted by the game at all, because the solver ranks by letter frequency, not by how likely a word is to be chosen. Over the original answer list in NYT's code, the same two guesses leave exactly one word: ELEGY. When the solver's list mixes everyday words with oddities, the everyday word is the better bet.
The next day's answer, SLOOP on 27 September, behaves the same way. SLATE gives green S and L; CRONY adds a green O in the middle. The solver lists nine words, led by SLOID and SLOJD, with SLOOP sixth. The original answer list allows only SLOSH and SLOOP, and the second O — a repeated letter — is what separates them.
The answer page for any date shows the word the Times actually chose, so the archive is also a way to replay old boards against the solver and see where the real answer ranked.
Repeated letters and the grey-tile rule
The single most common reason a hand-kept candidate list is wrong is a repeated letter. Suppose you guess ROBOT and the game colours the first O yellow and the second O grey. Most people cross O off, or keep it in an unclear way. The rule is precise: the answer contains exactly one O, and it is in neither position two nor position four. The solver encodes that exactly, so it drops double-O words and keeps single-O words with the O somewhere else.
The reverse case matters too. If you guess a word with one E and the answer has two, the game colours the one E you played and says nothing about a second. The solver knows there is at least one E, not exactly one, and keeps double-E words in the list. Players who reason "the E is placed, so no more Es" lose the answer whenever it repeats, and repeats are common: 747 of the 2,309 original answers have a repeated letter, and 10 of the 41 answers from September and early October 2026 did, PEEVE among them. Entering the colours tile by tile, rather than summarising them in your head, is the whole discipline the tool asks of you.
Hard mode, and why the solver's advice changes
Hard mode adds one rule, in the game's own words: any revealed hints must be used in subsequent guesses. That forbids probe words — you cannot play a word that ignores a known letter. The solver has no hard-mode switch because it does not need one: everything it lists is consistent with all the colours you entered, so every suggestion is a legal hard-mode guess.
The trap hard mode sets is the crowded ending. A green frame like _IGHT has nine answers in the original list — EIGHT, FIGHT, LIGHT, MIGHT, NIGHT, RIGHT, SIGHT, TIGHT, WIGHT. In normal mode a probe that tests several of those first letters at once can settle it; in hard mode every guess must end in IGHT, so each one tests a single first letter. The solver makes the situation visible by showing a list of near-identical words that no single guess from the list can split, which is the signal to choose the most common word and accept the odds.
Frequently asked questions
Why does the solver suggest words I've never heard of?
Its dictionary is a broad public-domain list of 15,910 five-letter words, and it ranks by common letters, not by how likely the Times is to choose a word. Some of its suggestions are not even accepted by the game; prefer the everyday word when two are close.
Does it work in hard mode?
Yes, without a switch. Every word it lists is consistent with all the greens and yellows you entered, which is what hard mode requires.
Is the tool's word list the official one?
No. It is the dwyl english-words list filtered to five letters. It overlaps the Times' accepted list heavily but not completely, and it lacks nine words from the original answer list, such as LATTE and RAMEN.