Tool
Algorithm Comparator
Pick up to five algorithms and see specifications, security bits, and a live throughput benchmark side-by-side. Useful for “which hash should I pick for X” decisions.
Pick up to 5 algorithms to compare:
| Property | MD5 | SHA-256 | SHA3-256 | BLAKE3-256 |
|---|---|---|---|---|
| Output bits | 128 | 256 | 256 | 256 |
| Block size (bits) | 512 | 512 | 1088 | 1024 |
| Rounds | 64 | 64 | 24 | 7 |
| Construction | Merkle-Damgård + Davies-Meyer | Merkle-Damgård + Davies-Meyer | Keccak sponge | Merkle tree of BLAKE2-style compression |
| Year | 1992 | 2001 | 2015 | 2020 |
| Designer | Ronald Rivest | NSA | Bertoni, Daemen, Peeters, Van Assche | O'Connor, Aumasson, Neves, Wilcox-O'Hearn |
| Standard | RFC 1321 | NIST FIPS 180-4 | NIST FIPS 202 | , |
| Collision sec. | 2^18 | 2^128 | 2^128 | 2^128 |
| Preimage sec. | 2^123 | 2^256 | 2^256 | 2^256 |
| Length-extension | yes (avoid raw MAC) | yes (avoid raw MAC) | no | no |
| Status | broken | ok | modern | modern |
| Primary uses | Legacy file checksums, Historical TLS / signatures | TLS, Bitcoin (SHA-256d), Git (transitioning), HMAC-SHA-256, JWT HS256 | Ethereum (note: uses Keccak-256, not SHA3-256 final), TLS optional | File integrity at scale, Build systems, Content-addressing |
| One-liner | First widely-used 128-bit hash; collisions trivial today. | The workhorse hash. Vulnerable to length-extension when used naively. | Drop-in alternative to SHA-256 with a different construction. | Parallel, tree-based; the fastest secure hash on modern CPUs. |
| Throughput 4 MiB | run benchmark | run benchmark | run benchmark | run benchmark |
Times are wall-clock in this browser tab. Two passes are run; the first is discarded as warm-up.
How to read this
- Collision security follows the birthday bound: output_bits / 2. A status of broken reflects a real-world attack better than this bound.
- Length-extension is a property of the construction. Merkle-Damgård hashes leak it; sponge (SHA-3), HAIFA (BLAKE2/3), and truncated MD variants (SHA-512/256, SHA-384) do not.
- Throughput is measured in this browser tab on a buffer of pseudo-random data. SIMD-using algorithms (BLAKE3, BLAKE2b) win big on capable CPUs; smaller hashes (MD5) finish fast simply because each block does less work.
FAQ