UUID Generator (v4 and v7)

UUID format and versions

RFC 9562 (2024) is the current UUID standard; it replaced RFC 4122. The 128 bits are written as hex in groups of 8-4-4-4-12. The first digit of the third group is the version, and the first digit of the fourth group shows the variant: 8, 9, a or b for standard UUIDs. Uppercase and lowercase mean the same ID, though most systems print lowercase.

• v1: a timestamp plus a node ID, often the MAC address, so it shows when and where it was made.
• v3 and v5: a hash of a namespace and a name (MD5 or SHA-1). The same name always gives the same UUID.
• v4: 122 random bits. The most common version, and what crypto.randomUUID() returns.
• v6: v1 with the time bits reordered so it sorts by time.
• v7: a 48-bit Unix time in milliseconds, then 74 random bits. It sorts by creation time; RFC 9562 recommends it over v1 and v6.
• v8: a free layout for vendor-specific IDs.
• Nil (00000000-0000-0000-0000-000000000000) and Max (ffffffff-ffff-ffff-ffff-ffffffffffff): special values, often meaning "no ID" or "unset".

UUID v4 or v7?

Pick v7 for database primary keys. A v4 lands at a random spot in a B-tree index, so inserts touch pages across the whole index and slow down as the table grows. A v7 starts with the time, so new rows go to the end of the index, like an auto-increment number.

Pick v4 when the ID must not reveal anything. Anyone can read a v7's creation time, as Check a UUID shows, so a v7 in a public URL tells people when an account or order was created.

The v7 values here are random after the timestamp, so IDs from the same millisecond aren't in creation order. RFC 9562 allows this. For strict order inside one process, use a library that adds a counter (RFC 9562, section 6.2).

Name-based UUIDs (v3 and v5)

When the same input must always give the same ID, for example one ID per URL across several services, use v5. Python's uuid.uuid5(uuid.NAMESPACE_DNS, "www.example.com") always returns 2ed6657d-e927-568b-95e1-2665a8aea6a2, the test value from RFC 9562, which prefers v5 over v3. The generator here doesn't create v3 or v5, but Check a UUID reads their version.

Validate a UUID

To accept only standard UUIDs of versions 1 to 8, match ^[0-9a-f]{8}-[0-9a-f]{4}-[1-8][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$ without case sensitivity. It rejects braces, missing hyphens and the Nil and Max values. Try it on your own data in the Regex Tester.

Storing UUIDs in a database

Store UUIDs as 16 bytes, not as 36 characters of text. PostgreSQL has a native uuid type. In MySQL, use BINARY(16) with UUID_TO_BIN() and BIN_TO_UUID(); MariaDB 10.7 and later has its own UUID type. SQL Server's uniqueidentifier sorts by the last group first, so a v7 there doesn't keep inserts in order; NEWSEQUENTIALID() is the SQL Server way to get that.

Generate UUIDs in code

Use the built-in function or a well-known library. Don't build UUIDs from Math.random().

crypto.randomUUID(); // v4, browsers and Node.js 19+

import uuid

uuid.uuid4()
uuid.uuid7()  # Python 3.14+
uuid.uuid5(uuid.NAMESPACE_DNS, "www.example.com")

id := uuid.New()         // v4
id7, err := uuid.NewV7() // v7

UUID.randomUUID(); // v4

Guid.NewGuid();         // v4
Guid.CreateVersion7();  // v7, .NET 9+

Uuid::new_v4();
Uuid::now_v7();

require "securerandom"
SecureRandom.uuid # v4

Ramsey\Uuid\Uuid::uuid4()->toString();

SELECT gen_random_uuid();  -- v4, PostgreSQL 13+
SELECT uuidv7();           -- v7, PostgreSQL 18+

SELECT UUID();                  -- v1
SELECT UUID_TO_BIN(UUID(), 1);  -- 16 bytes, time parts moved first

uuidgen
cat /proc/sys/kernel/random/uuid  # Linux, v4