Password Strength Tester
Password strength meter. Live entropy calculation + crack-time estimation at 4 attack speeds (online throttled, online unthrottled, offline slow hash, offline fast hash). Detects common passwords, sequential runs, keyboard patterns. Password never leaves the page.
Password Strength Tester
How the password test works
Type or paste a password
The result updates as you type. Press "Show" to check what you typed. Everything happens in your browser: the password is never sent, stored or logged, and this page does not look it up in any breach database.
Read the two entropy figures
The first is what the password would be worth if every character had been picked at random from the kinds of character it uses. That is a ceiling, not a measurement. The second takes off what this page can recognise: common passwords, a word followed by digits, runs like "abc" or "qwerty" and repeated characters. The rating uses the second figure: under 28 bits Very weak, then Weak, Fair from 36, Strong from 60 and Very strong from 80. Those bands are this page's own.
Check the attack times
Five rows show how long it would take to try every candidate. The first four are the scenarios of the zxcvbn estimator: an online login that allows 100 guesses an hour, one that allows 10 a second, a stolen database protected by a slow hash such as bcrypt (10,000 a second) and one protected by a fast hash (10 billion a second). The fifth is one RTX 4090 graphics card attacking unsalted MD5 at 164.1 billion a second, from hashcat's published benchmark.
Fix what it finds, then lengthen
Weaknesses the page spots are listed under the times. The page also checks length against NIST's current minimums: 15 characters for a password used on its own, 8 when it is one of two factors.
What a strength meter can and cannot tell you
Have I Been Pwned lists more than 17.8 billion breached accounts across 1,045 breaches (its public breach list, read 16 October 2026). Once a site's password database leaks, the attacker no longer has to guess through a login form. They can test guesses against the stolen hashes as fast as their hardware allows, and a password's strength is then the number of guesses it takes to find.
Entropy only measures randomness
Entropy in bits is the base-2 logarithm of how many equally likely passwords could have been chosen. For a password picked at random it is exact: 12 characters drawn from all 95 printable ASCII characters carry 12 × log₂95 = 78.8 bits. For a password a person made up, it is not, because people do not choose at random. NIST's guidance says as much.
"…estimating entropy for user-chosen passwords is challenging."NIST SP 800-63B, Appendix A
That is why this page shows two figures. Password1! uses all four kinds of character, so as random characters it would be worth 65.7 bits. This page recognises it as a common word plus a digit and caps it at 16 bits, about 65,000 guesses, which a fast-hash attack covers in well under a second. The cap is a rough allowance, not a count of what any real cracking tool would need.
The opposite mistake happens too. tilt-purple-fjord-1894 sees no pattern it knows, so the meter rates it as 22 random characters, 134.4 bits. If it was made by picking three words at random from the EFF's 7,776-word list and four random digits, it is worth 3 × log₂7,776 + log₂10,000 ≈ 52.1 bits: still good, but far less than the meter says. If you chose the words yourself, it is worth less again. The meter cannot tell how a password was made, only what it looks like.
Why the attack rate matters more than the password
The same password can last centuries or a second depending on how the site stored it. In hashcat's benchmark one RTX 4090 runs 164.1 billion MD5 guesses a second but only 184,000 bcrypt guesses at cost 5, and bcrypt's cost is a power of two, so at cost 10 that falls to about 5,750 a second. Online, NIST requires services to disable a password after no more than 100 consecutive failed attempts, which is why a throttled login is the slowest row.
Privacy
The check runs entirely in your browser. The password is not sent, stored or logged, and the page does not look it up in Have I Been Pwned or any other breach list. The page itself still makes its own requests, such as its scripts, an anonymous performance measurement, and ads or analytics only if you allow them; none of them carries what you type into the field.
MD5 falls at billions of guesses a second; bcrypt's cost factor is what stops that
Have I Been Pwned lists more than 17.8 billion breached accounts across 1,045 breaches.
The UK's NCSC found 23.2 million breached accounts using "123456" in its 2019 analysis of breach data.
One RTX 4090 tries 164.1 billion MD5 hashes a second in hashcat's benchmark. Eight random lowercase letters last about a second.
At that rate, 12 random printable ASCII characters would still take about 104 thousand years to exhaust.
Each step of bcrypt's cost factor doubles the work: the same card manages 184,000 a second at cost 5 and about 5,750 at cost 10.
NIST now requires 15 characters for a password used on its own, and 8 when it is one of two factors.
NIST says services must not force periodic password changes or impose rules mixing character types.
Four words picked at random from the EFF's 7,776-word list carry 51.7 bits; five carry 64.6.
NIST requires services to allow password managers and autofill, and to check new passwords against a blocklist of breached and common ones.
Length beats variety: 20 random lowercase letters (94.0 bits) outlast 8 random characters from all 95 printable ones (52.6 bits).
Frequently asked questions
No. The check runs in your browser and the password is never uploaded, stored or logged. The field is not part of any form, so nothing can submit it. The page does make its own requests, for its scripts, an anonymous performance measurement, and ads or analytics only if you allow them, but none of them carries the password.
No. It does not contact Have I Been Pwned or any other breach list, so a good rating here does not mean the password has never leaked. NIST requires services to check new passwords against a blocklist of breached and common ones; this page's own list has only 120 entries.
A count of how many equally likely passwords could have been chosen, expressed in bits; each extra bit doubles the number of guesses. It is exact only for passwords chosen at random. For a password a person made up it is an upper limit, which is why this page also shows a figure after the patterns it can spot.
Because attackers try common words with a capital letter and a digit or two added before anything else. As random characters it would be worth 65.7 bits; this page recognises the word-plus-digit shape and caps it at 16 bits, about 65,000 guesses.
Only when there is reason to think one has been exposed. NIST's current guidance says services must not require periodic changes, but must force a change when there is evidence of compromise.
An online attacker has to go through the login page, and NIST requires services to disable the password after no more than 100 consecutive failures. An offline attacker has a copy of the stored hashes and can test them on their own hardware as fast as it runs, which for fast hashes such as MD5 is billions a second.
They can be, if the words really are picked at random. From the EFF's list of 7,776 words, each word adds 12.9 bits: four words give 51.7 bits and five give 64.6. Words you choose yourself are worth much less, and this meter cannot tell the difference.
A little. Going from 26 possible characters to 95 raises each character from 4.7 to 6.6 bits. Adding characters helps more: going from 8 to 16 random characters doubles the total.
Because an attacker guesses whole words, not letters. Even a 200,000-word dictionary gives a single word only about 17.6 bits, however long the word is. This page knows only a short list, so it will rate many real words higher than an attacker would.
Yes. NIST requires services to allow password managers and autofill, and notes that managers with built-in generators make people more likely to choose strong passwords. Most browsers include one, and dedicated managers are also available.
Method & sources
What this tool implements
- NIST SP 800-63B (800-63-4 revision): 15 characters for a single-factor password, 8 when used only within multi-factor authentication; each Unicode code point counts as one character; no composition rules; no periodic changes, but a forced change on evidence of compromise; check new passwords against a blocklist; allow password managers and paste.
- Entropy is shown twice and labelled: the bound if every character were random, and the estimate after the patterns this page can spot. NIST: estimating entropy for user-chosen passwords is challenging. The rating bands (28/36/60/80 bits) are this page's own.
- Attack rates are zxcvbn's published scenarios plus one measured benchmark: hashcat 6.2.6 on one RTX 4090, MD5 164.1 GH/s and bcrypt 184.0 kH/s at cost 5 (about 5,750 a second at cost 10). Times are for the whole space; the average is half.
- Corrected 2026-10-16: 10 guesses/s labelled "throttled" and an invented 100/s row (zxcvbn: 100/hour and 10/s); "1k years" for 101 years; emoji counted as two characters; "RTX 4090 makes any sub-12-character password instantly crackable" (95^12 takes about 104 thousand years at 164.1 GH/s); HIBP "tens of billions of records" (17.8 billion accounts, 1,045 breaches, 16 Oct 2026); 26→95 characters "doubles" per-character entropy (4.7→6.6 bits); a dictionary word "~12 bits" (log2 200,000 = 17.6); "Password1!" "~12 bits" (this page caps it at 16); a 4-word example passphrase "60+ bits" (about 52 if random words); "2FA defeats ~99%" (unsourced, removed); an unattributed pull-quote; "no further network calls after load".
Sources
- National Institute of Standards and Technology. NIST SP 800-63B, Digital Identity Guidelines: Authentication and Authenticator Management (800-63-4). https://pages.nist.gov/800-63-4/sp800-63b.html
- Dropbox. zxcvbn — attack-time scenarios and score thresholds (src/time_estimates.coffee). https://github.com/dropbox/zxcvbn/blob/master/src/time_estimates.…
- Chick3nman. hashcat v6.2.6 benchmark, NVIDIA GeForce RTX 4090 (MD5 164.1 GH/s; bcrypt $2*$ 32 iterations 184.0 kH/s). https://gist.github.com/Chick3nman/32e662a5bb63bc4f51b847bb422222fd
- Electronic Frontier Foundation. EFF large wordlist for passphrases (7,776 words). https://www.eff.org/files/2016/07/18/eff_large_wordlist.txt
- Hunt T. Have I Been Pwned — breach list (API v3 /breaches; 1,045 breaches, 17,846,363,019 accounts on 2026-10-16). https://haveibeenpwned.com/api/v3/breaches
- National Cyber Security Centre. Most hacked passwords revealed as UK cyber survey exposes gaps in online security (2019): 23.2 million breached accounts used 123456. https://www.ncsc.gov.uk/news/most-hacked-passwords-revealed-as-uk…
What can make this go out of date
- None at runtime; the password never leaves the browser and no breach service is called.
Pick up where you left off
Stored only in this browser — never sent to our servers.