UUID Versions Explained

**By Dev Tools Weekly** | createuuid.com
When to use v1 vs v4 vs v5 vs v7 — with code examples.

---

UUID Format

xxxxxxxx-xxxx-Vxxx-Yxxx-xxxxxxxxxxxx
              │    │
              │    └── Variant (8, 9, a, or b = RFC 4122)
              └─────── Version digit (1-8)
Total: 128 bits = 32 hex characters + 4 hyphens = 36 characters

Example: 550e8400-e29b-41d4-a716-446655440000

• Version: 4 (position 13)
• Variant: a (position 17, starts with binary 10xx)

---

Version Comparison

---

Deep Dive: Each Version

Version 1 — Timestamp + MAC Address

Structure:  time_low - time_mid - time_hi_version - clock_seq - node(MAC)
Bits:       32         16         16                 16          48

• **Timestamp:** 60-bit count of 100ns intervals since Oct 15, 1582
• **Clock sequence:** 14 bits, prevents duplicates if clock resets
• **Node:** 48-bit MAC address of the generating machine

⚠️ Privacy concern: Embeds the machine's real MAC address. Avoid in public-facing IDs.

✅ Good for: Internal systems where time-ordering matters and you control the infrastructure.

Version 4 — Random

All 122 bits are randomly generated (6 bits reserved for version/variant).

• **Collision probability:** ~50% chance of one duplicate after generating 2.71 × 10¹⁸ UUIDs
• **That's:** 1 billion UUIDs per second for 86 years before a 50% collision chance
• **No coordination needed** — any machine can generate independently

✅ Good for: Almost everything. Default choice for database PKs, API keys, session IDs.

Version 5 — Name-Based (SHA-1)

UUID = SHA-1(namespace_UUID + name)[first 128 bits]

Predefined namespaces:

Key property: Same namespace + same name → always the same UUID.

✅ Good for: Deterministic IDs from URLs, emails, domain names. Idempotent operations.

Version 7 — Unix Timestamp + Random (NEW — RFC 9562)

Structure:
┌──────────────────┬────┬──────────────────────────────┐
│  Unix ms (48 bit) │ V  │   Random (74 bits)           │
└──────────────────┴────┴──────────────────────────────┘

• **Timestamp:** 48-bit Unix millisecond epoch (good until year 10889)
• **Random:** 74 random bits for uniqueness
• **Sortable:** Sorts chronologically when compared as strings or bytes

✅ Best for databases: Natural sort order = better B-tree index performance than random v4. No MAC leak like v1.

---

UUID vs. Alternatives

---

Code Examples

JavaScript / TypeScript

// Native (v4 random)
crypto.randomUUID();
// → "550e8400-e29b-41d4-a716-446655440000"
// UUID v7 (using 'uuid' package)
import { v7 as uuidv7 } from 'uuid';
uuidv7();
// → "01902e1c-0e2d-7000-8a3b-2c1d4e5f6a7b"

Python

import uuid
uuid.uuid4()        # Random
uuid.uuid1()        # Timestamp + MAC
uuid.uuid5(uuid.NAMESPACE_DNS, "example.com")  # Name-based

PostgreSQL

-- Built-in (v4)
SELECT gen_random_uuid();
-- v7 (PostgreSQL 17+)
SELECT uuidv7();  -- or use extension
-- Use as primary key
CREATE TABLE users (
    id UUID DEFAULT gen_random_uuid() PRIMARY KEY,
    name TEXT NOT NULL
);

Go

import "github.com/google/uuid"
id := uuid.New()           // v4
id := uuid.Must(uuid.NewV7())  // v7

---

Decision Flowchart

Need a unique ID?
├── Need to be deterministic (same input → same ID)?
│   └── YES → Use v5 (name-based SHA-1)
├── Need time-sortable IDs?
│   ├── For a database PK? → Use v7
│   └── For distributed systems? → Use v7 or ULID
└── Just need random uniqueness?
    └── Use v4 (the default)

---

🛠 Generate UUIDs instantly: createuuid.com

More UUID guides: CreateUUID articles