The quick answer
Against an offline GPU attack (~10¹¹ guesses/second) on a fast hash: an 8-character random password falls in minutes, a 12-character one in centuries, and 16+ is effectively uncrackable. Against slow, modern hashing (bcrypt/Argon2) or throttled online logins, everything shifts massively in your favor. Human-picked passwords are weaker than every number on this page — the table assumes true randomness.
The time-to-crack table (offline, fast hash — worst case)
Average time to guess, assuming a perfectly random password and leaked unsalted MD5-style hashes at ~10¹¹ guesses/second:
| Characters | Digits only | Lowercase | Mixed A–z 0–9 | All types (95) |
|---|---|---|---|---|
| 8 | Instantly | Instantly | Instantly | ~2 minutes |
| 10 | Instantly | ~7 seconds | ~7 days | ~1.9 years |
| 12 | Instantly | ~5 hours | ~1,150 years | Millions of years |
| 14 | ~2 minutes | ~6 days | Millions of years | Beyond feasible |
| 16 | ~3 hours | ~1.5 years | Beyond feasible | Beyond feasible |
| 20 | Millions of years | Millions of years | Beyond feasible | Beyond feasible |
"Beyond feasible" means longer than any attacker's budget, patience, or civilization — often measured in universe-lifetimes.
The assumptions behind the numbers
- Offline attack on a fast hash — the "your database leaked and was hashed with MD5/SHA-1" scenario. This is the number that matters for random generation.
- Slow hashing flips the table. If the site used bcrypt/Argon2, the same rig manages ~10⁴–10⁵ guesses/second — a million-fold defense you don't control but benefit from.
- Online attacks are nearly irrelevant for length — rate-limited logins make even modest random passwords safe for centuries; that's what 2FA is for.
- Random generation is assumed. A human-typed "16-character password" usually contains words, names or patterns that collapse these numbers — see Common Password Mistakes.
The two moves that matter
- Go longer, not fancier. Every extra random character multiplies crack time by its pool size (26–95×).
- Be random per account. A century-strength password reused on a breached site is worth zero — get a manager to make both effortless.
Test your own
See the assumption-transparent estimate for any password on the checker (it never leaves your browser), or model exact pools and lengths with the entropy calculator. The deeper math is in How Password Entropy Works.
Frequently asked questions
How long does it take to crack a 12-character password?
A randomly generated 12-character password with all character types takes millions of years at 10¹¹ guesses/second offline. Mixed alphanumeric without symbols: ~1,150 years. Human-chosen 12-character passwords (words + substitutions) can fall in hours.
Do symbols significantly increase crack time?
The first symbol expands the pool from 62 to 95 characters (~1.5× crack time per character), which helps — but adding two more characters of any type multiplies time by the pool size twice over. Length first, variety second.
Can quantum computers crack my password instantly?
No. Grover's algorithm would square-root classical search spaces at best, on hardware that does not exist at relevant scale. Your 128-bit-entropy password is safe for the foreseeable future; passwords are not the weak link quantum computers threaten first.
Does the website's hashing really change it that much?
Yes — million-fold. MD5 allows ~10¹¹ guesses/second on a good rig; tuned Argon2id can limit the same rig to ~10⁴. You can't control a site's hashing, which is exactly why unique, random, long passwords are your side of the bargain.
Apply what you learned with our zero-telemetry tools and references:
Put it into practice: run your next password through the local checker — it never leaves your browser.