Tool
What the Brute-Force Time Estimator does
Brute-Force Time Estimator: calculates Entropy from Password to test. Example: password to test Tr0ub4dour&3 gives entropy 79 bits.
Give the Brute-Force Time Estimator your Password to test and it works out Entropy. Values update as you type, so comparing options is a matter of editing one field.
Typical users are someone verifying a download. Reusing a password is still the single most common way accounts fall over.
Because it is all local, the tool is as fast as your device and as private as your own notes.
The field table, the method and a worked example are documented after the tool for anyone who wants to check the working.
What do the Brute-Force Time Estimator fields mean?
The Brute-Force Time Estimator uses 1 input. Nothing is blank on load, so you can see the shape of the answer before supplying your own figures.
| Field | What to enter | Default |
|---|---|---|
| Password to test | required | Tr0ub4dour&3 |
How does the Brute-Force Time Estimator work?
Each output is derived from the inputs above in a single pass; there is no hidden state carried between runs, so the same inputs always give the same calculation.
Bad input produces an error, never a confident-looking wrong answer.
Worked example
These are the values the Brute-Force Time Estimator loads by default, and the result it produces from them. These are the exact values loaded when the page opens, and the answer they produce.
| Password to test | Tr0ub4dour&3 |
|---|
| Entropy | 79 bits |
|---|
Why the same password has four different crack times
Time to crack is not a property of the password alone. It depends entirely on the attack scenario, and the four scenarios this tool compares span roughly nine orders of magnitude in guess rate.
| Online, rate-limited | ~100 guesses/sec | Most login forms after a handful of failed attempts |
|---|---|---|
| Online, unthrottled | ~1 million guesses/sec | A poorly defended API or login endpoint with no lockout |
| Offline, slow hash (bcrypt) | ~10 thousand guesses/sec | A leaked database where passwords were hashed correctly |
| Offline, fast hash / GPU rig | ~100 billion guesses/sec | A leaked database hashed with a bare SHA-256, or worse |
Worth knowing
- These are commonly cited order-of-magnitude figures for a single modern GPU rig, not measured benchmarks. Real throughput depends on the exact hash, the hardware, and how many GPUs an attacker has, and can easily move by 10× in either direction.
- The nine-order-of-magnitude spread means the single most consequential decision is usually not how a user chose their password. It is how the service stored it. A weak hashing choice makes every user’s password weaker at once, by the same enormous factor.
- Rate limiting and account lockouts are why online attacks rarely matter in practice; almost any non-trivial password survives a few thousand throttled attempts. Offline cracking after a breach is the scenario that actually decides outcomes.
- See the hash generator and HMAC generator entries below for the mechanics of why bcrypt is deliberately thousands of times slower than SHA-256 at the same nominal task.
Read next: What Actually Makes a Password Strong
How to use it
- Set Password to test.
- Results update continuously rather than on submit.
- Check Entropy on the right.
- Print gives a tidy version without the navigation, which is handy for attaching to a file.
What this tool does not do
- Randomness comes from your browser's cryptographic random source, which is suitable for passwords and keys.
- Anything generated here is only as safe as where you store it afterwards.
- Treat it as a fast first pass, not as the final word on a decision that matters.
Frequently asked questions
You need Password to test. Nothing starts empty, so nothing is mandatory before you see a result.
With the values loaded when the page opens, entropy comes out as 79 bits. Overwrite any of the starting values and the result follows.
No. There is no account, no session and no storage. The values exist only while the tab is open.
No payment and no login. Advertising covers the running costs.
The computation is exact; how much you should trust the answer depends on how good your inputs are.