Popular tools

Theme

Password Strength Tester

See how long your password would take to crack, attack by attack, which patterns give it away and whether it has appeared in a data breach. This password strength tester is a password checker that never sends your password: only 5 characters of its hash leave your device.

Pattern estimate: zxcvbn-ts, corrected · attack speeds: hashcat benchmarks · breach data: Have I Been Pwned · nothing stored
Checked in this browser
Every part an attacker would guess shows up here:
  • Common password
  • Word
  • Name
  • Year
  • Keyboard
  • Repeat
  • Random

Tested in your browser. Only the first 5 of the 40 characters of its SHA-1 hash leave it, for the breach check.

Try

How fast would it fall?

Type or paste a password: the verdict updates as you type.

Bits an attacker must guess
0128

What it checks

  • Patterns attackers try first: 49,233 common passwords, words and names, years, keyboard runs, repeats and look-alike swaps.
  • Data breaches: about 2.07 billion leaked passwords in Have I Been Pwned, by k-anonymity.
  • Eight attacks, from a login page that locks after 100 tries to NTLM hashes on 100 graphics cards.

How strong is my password?

As strong as the number of guesses an attacker needs to reach it, and attackers try likely passwords first: common passwords, words and names, years, keyboard runs and swapped letters, long before random strings. The tester counts those guesses.

It splits what you type into the pieces attackers try, such as Summer + 2026 + !, prices each piece by how many guesses it takes, and adds up the cheapest way to build the whole password. The engine is zxcvbn-ts, the open-source estimator OWASP points to, with English word lists and those of the page's language where one exists.

We corrected the two places where the stock estimator goes wrong at the strong end: it rates a truly random 12-character password at about 10¹² guesses instead of about 10²⁴, and a six-word passphrase billions of times too high. Random stretches are now priced by the characters they use, and passphrases by the scheme that made them.

Capitals, digits and symbols are shown, but a missing type never counts against you: NIST dropped such rules because they produce predictable passwords like Password1! (NIST SP 800-63B-4). The first 64 characters are analysed; anything longer is beyond guessing unless it's a known phrase, and the breach check still sees all of it.

Summer2026!

  1. SummerDictionary wordA season, capitalised the usual way: 50 guesses
  2. 2026!RandomFive characters with no pattern found: 100,000 guesses
0128

WeakAbout 10 million guesses: WeakExposedIn breach data 45 times: Exposed

The patterns, not the character types, decide: 11 characters of four types, guessed in seconds once a list of hashes leaks.
Password strength tester: a password split into the patterns that weaken it, crack times for 8 attacks and a breach checkPassword strength tester: a password split into the patterns that weaken it, crack times for 8 attacks and a breach checkThe ratingVery weak to Very strong, from the guesses an attacker needs, or Exposed when the password is in breach data.Where it's weakThe password split into the patterns attackers try, each with the reason it's cheap to guess.Crack timeHow long each of eight attacks would take, on one graphics card or a hundred.Breach checkWhether Have I Been Pwned has seen it, and the 5 characters of its hash that were sent.
A made-up password tested: the rating, the patterns that give it away, the time each attack would take and the breach check. Select a number to see what each part shows.

How long would it take to crack my password?

It depends on the attack: the same password can hold out for years against a login page that limits tries and fall in under a second once a site leaks badly stored password hashes. The tester works as a password crack time calculator for eight named attacks, on one graphics card or a hundred.

How long to crack a password comes down to the guesses it needs divided by the attacker's speed. Online, attackers guess through the login page: 100 tries an hour where the site locks out repeated failures (NIST allows at most 100 in a row), 10 a second where it doesn't. Offline, after a database leak, they guess against the stored hashes on graphics cards, and the speed depends on how the site stored them.

One RTX 5090 tries 340.1 billion NTLM hashes a second (Windows), 220.6 billion MD5 and 70.2 billion SHA-1, but about 9,500 bcrypt hashes at cost 10 and 1.48 million Linux sha512crypt (hashcat benchmarks). Argon2id with 64 MiB of memory runs at 1,699 a second on an RTX 4090 (hashcat). That gap is why OWASP asks sites to store passwords with Argon2id or bcrypt.

Each time is an average: for a random password, half of all the possibilities. A password strength test is only as honest as the attack it assumes, so each row names one; real attackers also run lists and rules before brute force, which is what the pattern estimate models. Phishing and keyloggers skip guessing altogether.

Online, tries limitedOnline, no limitLeaked, Argon2id*Leaked, bcryptLeaked, Linux sha512cryptLeaked, SHA-1Leaked, MD5Leaked, NTLM
6-digit PIN105.7208 days14 hours5 minutes52 secondsless than a secondless than a secondless than a secondless than a second
Summer2026!, by its patterns alone107.011 years12 days2 hours17 minutes7 secondsless than a secondless than a secondless than a second
8 random characters1015.12B years4M years26K years5K years30 years6 hours2 hours1 hour
4 random words1015.32B years6M years34K years6K years39 years7 hours2 hours1 hour
12 random characters1023.210¹⁷ years10¹⁴ years10¹² years554B years4B years75K years24K years16K years
6 random words1023.010¹⁷ years10¹⁴ years10¹² years368B years2B years50K years16K years10K years
16 random characters1031.210²⁵ years10²² years10²⁰ years10¹⁹ years10¹⁷ years10¹² years10¹² years10¹² years
Any password in breach data109.32K years7 years14 days3 days23 minutesless than a secondless than a secondless than a second
Online, tries limitedOnline, no limitLeaked, Argon2id*Leaked, bcryptLeaked, Linux sha512cryptLeaked, SHA-1Leaked, MD5Leaked, NTLM
6-digit PIN105.7208 days14 hours3 secondsless than a secondless than a secondless than a secondless than a secondless than a second
Summer2026!, by its patterns alone107.011 years12 days59 seconds10 secondsless than a secondless than a secondless than a secondless than a second
8 random characters1015.12B years4M years262 years47 years110 days3 minutes1 minute41 seconds
4 random words1015.32B years6M years341 years61 years143 days4 minutes1 minute54 seconds
12 random characters1023.210¹⁷ years10¹⁴ years31B years6B years36M years752 years239 years155 years
6 random words1023.010¹⁷ years10¹⁴ years21B years4B years24M years499 years159 years103 years
16 random characters1031.210²⁵ years10²² years10¹⁸ years10¹⁷ years10¹⁵ years69B years22B years14B years
Any password in breach data109.32K years7 years3 hours36 minutes14 secondsless than a secondless than a secondless than a second
Average times. Speeds: hashcat benchmarks on one RTX 5090; Argon2id on an RTX 4090. Real attacks try lists and rules before brute force. * Argon2id: measured on an RTX 4090 (no RTX 5090 benchmark published yet).

Is my password secure? What each rating means

Your password is secure against guessing if it rates Very strong, or Strong at a site that stores passwords with a slow hash, and has never appeared in a breach. Exposed overrides everything else.

Each rating is a range of guesses, set by how long real attacks take to cover it. Very strong needs 10²¹ guesses or more: no single graphics card finds it within a century, even against NTLM.

Guesses an attacker needs
Exposed
  1. 6-digit PIN
  2. Summer2026!
  3. 8 random characters
  4. 4 random words
  5. 6 random words
  6. 12 random characters
  7. 16 random characters
Very weakWeakFairStrongVery strong
  1. Locked login
  2. Open login
  3. Argon2id
  4. bcrypt
  5. sha512crypt
  6. SHA-1
  7. MD5
  8. NTLM

▲: one year on one graphics card

Exposed sits outside the scale: a breached password is on attackers' lists, whatever its length.
RatingGuesses neededWhat it withstands
ExposedIn breach dataNothing: attackers try breached passwords first. Change it everywhere you use it.
Very weakUnder 10⁴Online guessing finds it within 17 minutes, even at 10 tries a second.
Weak10⁴ to 10⁸A site that locks out after 100 failures, but not one without a limit; after a leak, 16 hours at most.
Fair10⁸ to 10¹⁴Online guessing for months to millennia; after a leak, hours to 333 years against bcrypt, but under 10 minutes against MD5 or NTLM.
Strong10¹⁴ to 10²¹After a leak: 333 to 3 billion years against bcrypt, but 5 minutes to 93 years against NTLM, on one graphics card.
Very strong10²¹ or moreNo single graphics card finds it within a century, even against NTLM, the fastest hash to crack.

What makes a strong password?

Length and randomness: at least 15 characters if the password alone protects the account, made by a generator or from random words, used for one account only and never found in a breach. Symbols, capitals and regular changes are no longer required.

That is the rule set of NIST SP 800-63B-4, the US standard revised in July 2025: 15 characters single-factor and 8 with a second factor, at least 64 accepted, any characters including spaces and emoji, no composition rules, no forced changes, and every new password checked against breached and common ones.

CISA, the US cybersecurity agency, asks for at least 16 characters or a passphrase of four to seven unrelated words, and the UK's NCSC suggests three random words. All three put length first.

Swapping letters for look-alike digits doesn't help: attackers know o → 0 and a → @, and try them early (NCSC). The password generator makes passwords that meet every rule.

NIST SP 800-63B-4what the standard says · what the tester does
  • At least 15 characters, or 8 with a second factorLength checked against 15
  • Accept at least 64 charactersUp to 256 accepted, the first 64 analysed
  • Any characters, including spaces and emojiCounted in characters: an emoji is one
  • No rules such as "must contain a symbol"A missing type is never marked as a problem
  • No forced changes; change after a compromiseExposed says when it's time
  • Block common and breached passwordsChecked against Have I Been Pwned
  • Allow paste and password managersPaste works and triggers the breach check
NIST SP 800-63B-4, section 3.1.1.2, the US standard for passwords since July 2025.

Password entropy: why P@ssw0rd and Tr0ub4dor&3 fail

Entropy counts the possible passwords an attacker would have to try, which is meaningful only for a password drawn at random. Counted by character types, P@ssw0rd scores about 52 bits and Tr0ub4dor&3 about 72; both are in breach data, thousands to millions of times.

A password entropy calculator multiplies the length by the bits of each character type present, as if every character had been picked at random. A person never does that: P@ssw0rd is the word "password" with two well-known swaps, and Have I Been Pwned has seen it 6,421,042 times. That is why NIST bases its rules on length: entropy can't be worked out for passwords people choose (NIST SP 800-63B-4, Appendix A).

Tr0ub4dor&3, xkcd's example of a hard-to-remember password, is a dictionary word with swapped letters, a symbol and a digit: about a million guesses once the swaps are undone, Weak on its own and Exposed in fact, at 3,196 sightings.

For a random password the two counts agree, and the password generator shows its exact bits. The tester reports guesses instead: 10ˣ guesses is 3.32 × x bits.

  • By character types
  • By patterns
  • In breach data
  1. P@ssw0rd52.6 bits5.0 bits6,421,042 times
  2. Tr0ub4dor&372.3 bits62.2 bits3,196 times
  3. 12 random characters78.8 bits78.8 bitsNot found
Character types assume every character was random. Only for the random password do the two counts agree.

Has my password been leaked? A private Have I Been Pwned check

The tester runs a Have I Been Pwned password check on the exact password, against Pwned Passwords' about 2.07 billion passwords from data breaches, without sending it: only the first 5 of the 40 characters of its SHA-1 hash leave your browser, and they go to our server, not to Have I Been Pwned.

Your browser hashes the password with SHA-1 and sends the first 5 characters, such as 21BD1. About 2,000 hashes in the database start with those 5; the reply lists all of them with their counts, and your browser looks for the other 35 characters in that list. This is k-anonymity, Have I Been Pwned's own design (HIBP): whoever sees the 5 characters can't tell which of the 2,000 you hold.

Our server asks Have I Been Pwned on your behalf, so it never sees your IP address, pads each reply with 100 to 300 dummy rows so the size gives nothing away, and doesn't log the prefix. The check runs when you leave the field, press Enter or paste, never on each keystroke: a prefix per letter typed could give the password away, which is why Have I Been Pwned advises against it (HIBP).

The count is sightings, not breaches: Tr0ub4dor&3 was seen 3,196 times. The 2.07 billion is our own measurement, as Have I Been Pwned publishes no current total: 256 random prefixes held 1,971 hashes each on average (23 October 2026).

Your browser

••••••••

SHA-1 of the password

21BD12DC183F740EE76F27B78EB39C8AD972A757

Sent: the first 5 characters · Kept: the other 35

Compared in your browser

…2DC183F74…:6,421,042

Seen 6,421,042 times

dnsrobot.net

21BD1

+ 100 to 300 dummy rows

Not logged

About 2,000 hashes that start with 21BD1

Have I Been Pwned

/range/21BD1

Sees our server, never you

k-anonymity: anyone who sees the 5 characters can't tell which of about 2,000 passwords you hold.

What to do if your password was in a breach

Stop using it: change it first on your email account, then on every site where you used the same or a similar password, each time to a new random one, and turn on two-step sign-in wherever it's offered.

A breached password is on the lists attackers try first, so its strength no longer counts: running the whole list of 2.07 billion takes one graphics card under a second against NTLM or MD5 hashes and about 2.5 days against bcrypt. That is why NIST requires sites to reject breached passwords and to make you change one that has been compromised (NIST SP 800-63B-4).

Email comes first because it can reset every other account. The password generator gives each account its own password, and a password manager remembers them. A hit doesn't mean your account was broken into: the list holds the password itself, from whoever used it.

Even a strong, unique password doesn't stop phishing or a keylogger, which capture it as you type (NIST SP 800-63B-4). Check where a link really leads before you sign in, and check a suspicious email's real sender with the email header analyzer.

  1. 1Change it on your email firstEmail can reset every other account.
  2. 2Then everywhere you reused itEach account gets its own generated password.
  3. 3Turn on two-step sign-inA stolen password alone no longer gets in.
  4. 4Keep them in a password managerIt remembers each one and fills it in.

Password strength: questions and answers

Is it safe to type my real password here?

The password stays in your browser. The estimate runs on your device, and the breach check sends only the first 5 of the 40 characters of its SHA-1 hash, once you leave the field, press Enter or paste. Nothing is stored or logged, and spellcheck is off for the field. If you'd still rather not, test a password built the same way.

Is this the same as Have I Been Pwned's password check?

It uses the same data, Have I Been Pwned's Pwned Passwords, and the same k-anonymity range search. Three things differ: the request goes through our server, so Have I Been Pwned never sees your IP address; it runs once you finish typing, never per keystroke; and a hit changes the rating to Exposed instead of sitting beside a strength score.

What does "seen 3,196 times" mean?

It's how often that exact password appears in Have I Been Pwned's breach data: 3,196 sightings for Tr0ub4dor&3. It isn't the number of breaches or of accounts, and it doesn't mean your account was hacked: the list holds the password, not who used it. Any count above zero puts a password on the lists attackers try first.

How does the password checker work out the crack time?

It divides the guesses an attacker would need on average by each attack's speed. The guesses come from the pattern estimate; the speeds are published hashcat benchmarks for one RTX 5090 (an RTX 4090 for Argon2id), and 100 graphics cards divide the time by 100. Real attackers try lists and rules before brute force, which the pattern estimate models.

Why does another password strength checker give a different answer?

Most count character types and don't say which attack they assume. When we tried nine popular checkers on 23 October 2026, six presented Tr0ub4dor&3 as safe, with crack times from 4 months to 2,000 years, although it's in breach data 3,196 times. This one names every attack, prices the patterns attackers try first and lets a breach override the rating.

Is this a password entropy calculator?

In effect, yes: it estimates how many guesses a password needs, and 10ˣ guesses is about 3.32 × x bits. Unlike a password entropy calculator that counts character types, it looks for patterns first: by types, P@ssw0rd scores about 52 bits, yet it's in breach data 6.4 million times. For a random password the two agree, and the generator shows its exact bits.

Why is my long password rated weak?

Length helps only when the characters are unpredictable. Repeated letters, keyboard runs such as qwertyuiop, a famous phrase or a line from a song are among the first things attackers try: aaaaaaaaaaaaaaaa has 16 characters and rates Very weak. A long password can also be in breach data, which rates it Exposed: correcthorsebatterystaple appears 4,173 times.

Do I need symbols, numbers and capital letters?

Not for strength. NIST's current rules tell sites to stop demanding them, because they lead to predictable choices like Password1!, and the tester never marks a missing type as a problem. Length and randomness matter far more: 20 random lowercase letters (94 bits) beat 12 random characters of every type (78 bits). Some sites still insist, so the generator can add them.

How long does it take to crack an 8-character password?

A random one, from all 94 keyboard characters: about 2 hours on one RTX 5090 if the site stored it as an NTLM hash, 4 hours as MD5, and some 10,000 years with bcrypt at cost 10. An 8-character password a person chose, such as a word and a year, falls in under a second against those fast hashes and within hours against bcrypt.

Can a strong password still be hacked?

Yes, because strength only measures guessing. Phishing pages, keyloggers and malware capture a password as you type it, however long it is, and a password reused on a breached site gets tried on your other accounts. Use a different password for every site, turn on two-step sign-in, and check where a link really leads before you sign in.

Why does a password that looks strong say Exposed?

Because it's in breach data, and attackers try those passwords before anything else. Running the whole list of about 2.07 billion takes one graphics card under a second against MD5 or NTLM hashes and a few days against bcrypt, so complexity no longer protects it. NIST requires sites to reject breached passwords for the same reason. Exposed overrides every other rating.

Why doesn't the breach check run as I type?

Because a check on every keystroke would send a new prefix for each letter, and that series of prefixes could reveal the password to anyone watching the traffic. Have I Been Pwned advises waiting until the whole password is entered, so the tester checks when you leave the field, press Enter or paste. The strength rating still updates as you type.

Missing something?Suggest a feature or a new tool

Related tools