BLAKE3
Parallelizable hashing that beats every other secure algorithm on raw throughput. Has a built-in XOF, MAC, and KDF mode.
By Deepak Gupta ·
BLAKE3 is what BLAKE2 looks like if you redesign for SIMD and many-core CPUs. The internal tree structure means a 1 GB file's hash can be computed in parallel across every core on your laptop. On modern AMD/Intel hardware it routinely outruns SHA-256 even *with* SHA-NI acceleration. It's a single algorithm that fills multiple roles: plain hashing, keyed hashing (replaces HMAC), extendable-output (replaces SHAKE), and key derivation (replaces HKDF). The big trade-off is ecosystem maturity: SHA-256 is in every standard library forever; BLAKE3 still needs a dependency in most languages. Reach for it when raw throughput matters and you control the consumer.
Recommended uses
- ·Hashing huge files in parallel (backups, deduplication, content-addressed storage)
- ·Replacing HMAC + HKDF + SHAKE with a single primitive
- ·High-throughput Merkle-tree leaf hashing
Known attacks / caveats
- ·None practical.
Designed by
O'Connor, Aumasson, Neves, Wilcox-O'Hearn, published 2020.
Frequently asked questions
- Is BLAKE3 secure in 2026?
- Yes. BLAKE3 has no practical breaks as of 2026, and no attack better than brute force is known against it.
- What is BLAKE3 used for?
- Hashing huge files in parallel (backups, deduplication, content-addressed storage). Replacing HMAC + HKDF + SHAKE with a single primitive. High-throughput Merkle-tree leaf hashing.
- How long is the output of BLAKE3?
- 256 bits, which is 32 bytes, or 64 characters when written as hexadecimal. The length is fixed regardless of how large the input is.
- Can BLAKE3 be used to hash passwords?
- No, and this is the most consequential mistake people make with it. BLAKE3 is a fast general-purpose hash, and fast is exactly the wrong property for passwords: it lets an attacker with a stolen database test billions of guesses. Use a purpose-built password hashing function such as Argon2id instead.
- Who created BLAKE3?
- O'Connor, Aumasson, Neves, Wilcox-O'Hearn, published in 2020.