UUID in SQL

SQL UUID Generator

Every major database can mint a UUID — but only one of them gives you a time-ordered one by default, and that choice is what decides how your index behaves.

PostgreSQL 18 uuidv7()MySQL UUID_TO_BINSQL Server NEWID()

Version

How many

Uppercase Hyphens Braces { }

Result

Generated locally with the Web Crypto API — nothing leaves your browser.

Generating UUIDs in SQL

PostgreSQL has had gen_random_uuid() since 13, and version 18 added uuidv7() together with uuidv4() as an alias — so a table can now default to a time-ordered key with no extension and no client code: DEFAULT uuidv7(). MySQL offers UUID(), which returns a v1 value, and UUID_TO_BIN(uuid, 1) to store 16 bytes with the timestamp field reordered so the index stays sequential. SQL Server has NEWID() for random GUIDs and NEWSEQUENTIALID(), which only works as a column default.
The storage decision matters as much as the generator. A char(36) or varchar(36) column costs 36 bytes per row and compares as a string; PostgreSQL's native uuid, MySQL's binary(16) and SQL Server's uniqueidentifier all cost 16 bytes and compare as fixed-size values. On a table with millions of rows that difference shows up in both disk and index depth.

How to generate

Press Generate UUID above to mint a value in this browser — nothing is sent anywhere.
Copy one value with Copy, or set How many and use Copy all for a fixture list.
Switch Version if your code needs v7 instead of the v4 default, and turn on Formatting to match your stack's conventions.

Use Cases

Table defaults — DEFAULT uuidv7() so inserts need no identifier from the application.
Migration from integer keys — backfill with generated values, then switch the column type.
Event and audit tables where a time-ordered key keeps writes appending.

How SQL UUID Generator Compares

uuidv7() / NEWSEQUENTIALID()	Time-ordered: inserts land at the end of the index
gen_random_uuid() / NEWID()	Random: unique, but writes scatter across the index
char(36)	Readable, 36 bytes, compared as text — prefer a native type

Code Examples

PostgreSQL

-- PostgreSQL 13+: random. PostgreSQL 18+: also uuidv7()/uuidv4()
SELECT gen_random_uuid();
SELECT uuidv7();

CREATE TABLE events (
  id         uuid PRIMARY KEY DEFAULT uuidv7(),  -- time-ordered
  payload    jsonb NOT NULL,
  created_at timestamptz NOT NULL DEFAULT now()
);

MySQL

-- UUID() returns a v1 value; store 16 bytes, reordered for the index
CREATE TABLE t (
  id BINARY(16) PRIMARY KEY DEFAULT (UUID_TO_BIN(UUID(), 1)),
  name VARCHAR(100)
);

SELECT BIN_TO_UUID(id, 1) FROM t;

-- SQL Server
-- CREATE TABLE t (id UNIQUEIDENTIFIER DEFAULT NEWID());
-- CREATE TABLE t (id UNIQUEIDENTIFIER DEFAULT NEWSEQUENTIALID());

Frequently Asked Questions

Which SQL function generates a UUID v7?

PostgreSQL 18's uuidv7(). MySQL and SQL Server have no v7 function — use UUID_TO_BIN(UUID(), 1) on MySQL, or generate the value in the application.

Should a UUID primary key be a string column?

No. Use PostgreSQL uuid, MySQL binary(16) or SQL Server uniqueidentifier — 16 bytes instead of 36, and compared as a fixed-size value.

Why is my random UUID primary key slow on insert?

Because the values are not ordered, so each insert lands in a different part of the B-tree and pages split. Switch to uuidv7() or NEWSEQUENTIALID() to make writes append.